2,3-Butanediol dehydration catalyzed by silica-supported alkali phosphates

Article abstract of DOI:10.1016/j.apcata.2018.08.015

Characterization of acid-base centers and catalytic dehydration of 2,3-butanediol (BDO) was performed over a wide range of silica-supported alkali phosphates (M_P/SiO₂; M = Na, K, Cs; M:P = 0.5–3 mol:mol). Selectivity to 1,3-butadiene (BD) and 3-butene-2-ol (3B2OL) formed by elimination correlates with the densities of conjugated acid-base pairs and increases in the order Na < K ≈ Cs. Selectivity to 2,3-epoxybutane (BTO) formed by dehydrative epoxidation increases with the M/P ratio in the order Na < K ≤ Cs. Methyl ethyl ketone (MEK) and isobutyraldyde (IBA) formed by the rearrangement are produced by isolated Br?nsted acid centers. Conjugated acid-base pairs are probably composed of the end groups of polyphosphates having both acid P-OH and base P–O^??M⁺ moieties. Isolated Br?nsted acid centers are probably silica grafted phosphoric acid molecules at low M/P and –PO(OH)₂ end groups of oligophosphates at M/P > 1.5. Deactivation rate increases with the increase of M/P ratio in order Na < K < Cs. Deactivation patterns imply that sites responsible for elimination are active in dehydrative epoxidation. Dehydration of 3B2OL smoothly proceeds to BD, but the catalysts deactivate faster compared to BDO dehydration.

Full text of DOI:10.1016/j.apcata.2018.08.015

Accepted Manuscript
Title: 2,3-Butanediol Dehydration Catalyzed by
Silica-Supported Alkali Phosphates
Author: Wooyoung Kim Wookyun Shin Kyoung Jun Lee
Yong Seok Cho Hyung Soon Kim Igor N. Filimonov
PII:
S0926-860X(18)30409-5
DOI:
Reference:
https://doi.org/doi:10.1016/j.apcata.2018.08.015
APCATA 16786
To appear in:
Applied Catalysis A: General
Received date:
Revised date:
Accepted date:
23 March 2018
17 July 2018
17 August 2018
Pleasecitethisarticleas:W. Kim, W. Shin, K.J. Lee, Y.S. Cho, H.S. Kim, I.N. Filimonov,
,3-Butanediol Dehydration Catalyzed by Silica-Supported Alkali Phosphates, Applied
2
Catalysis A, General (2018), https://doi.org/10.1016/j.apcata.2018.08.015
This is a PDF file of an unedited manuscript that has been accepted for publication.
As a service to our customers we are providing this early version of the manuscript.
The manuscript will undergo copyediting, typesetting, and review of the resulting proof
before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that
apply to the journal pertain.

2,3-BUTANEDIOL DEHYDRATION CATALYZED BY SILICA-SUPPORTED ALKALI
PHOSPHATES.
Wooyoung Kim, Wookyun Shin, Kyoung Jun Lee, Yong Seok Cho, Hyung Soon Kim, and Igor N.
Filimonov*
GS Caltex Corporation, R&D Center, 359, Expo-ro, Yuseong-gu, Daejeon, 34122, Rep. of Korea
ABSTRACT
Characterization of acid-base centers and catalytic dehydration of 2,3-butanediol (BDO) was
performed over a wide range of silica-supported alkali phosphates (M P/SiO ; M=Na, K, Cs;
_
2
M:P=0.5–3 mol:mol). Selectivity to 1,3-butadiene (BD) and 3-butene-2-ol (3B2OL) formed by
elimination correlates with the densities of conjugated acid-base pairs and increases in the order
Na<K≈Cs. Selectivity to 2,3-epoxybutane (BTO) formed by dehydrative epoxidation increases with
the M/P ratio in the order Na<K≤Cs. Methyl ethyl ketone (MEK) and isobutyraldyde (IBA) formed
by the rearrangement are produced by isolated Brønsted acid centers. Conjugated acid-base pairs
-
are probably composed of the end groups of polyphosphates having both acid P-OH and base P-O
+
∙
∙∙M moieties. Isolated Brønsted acid centers are probably silica grafted phosphoric acid molecules
at low M/P and -PO(OH) end groups of oligophosphates at M/P>1.5.
2
Deactivation rate increases with the increase of M/P ratio in order Na<K<Cs. Deactivation
patterns imply that sites responsible for elimination are active in dehydrative epoxidation.
Dehydration of 3B2OL smoothly proceeds to BD, but the catalysts deactivate faster compared to
BDO dehydration.
KEYWORDS: 2,3-butanediol dehydration; silica-supported alkali phosphates; 1,3-butadiene;
methyl ethyl ketone; 3-butene-2-ol; 2,3-epoxy butane; elimination/rearrangement; acid-base sites.
Corresponding Author
*
E-mail: FLM@gscaltex.com, infilimonov@gmail.com.
Page 1 of 39

1. Introduction
The worldwide trend on the environmental improvement and sustainable development leads
to the replacement of fossil-based source materials by renewable ones. Fermentation of a
carbohydrate feedstock to 2,3-butanediol (BDO) is one of the attractive strategies due to high titers
of BDO in the microbial broth [1,2]. BDO can be converted by dehydration to large scale consumer
substances like MEK and BD. A number of reviews on the catalytic dehydration of diols (including
BDO) was published recently [3–5] indicating a growing interest in the field of BDO conversion.
BDO vapor may be catalytically dehydrated by silica-supported sodium phosphates in the
fixed bed reactor to BD+3B2OL and IBA+MEK [6]. These catalysts have a good performance:
long-term stability and high selectivity to the products of elimination — BD and 3B2OL. They have
certain limitations like the necessity of dilution of the reactant by a carrier gas and a short contact
time to preserve a good carbon balance and reduce deactivation rate. Very short contact times and a
diluted BDO feed are important requirements for the other catalytic systems. According to [7],
under these conditions almost 80% selectivity to the elimination products at 10-15% of BDO
conversion was reached over amphoteric -alumina that normally provides mostly rearrangement
products.
Good correlation of the elimination rate with the density of acid-base pairs of sodium
phosphate [6] implies that E1cB mechanism is realized. The latter requires concerted action of acid
and base centers in the ensemble with a reactant i.e., certain spatial organization of three entities.
Though the mechanisms of eliminations are a basic knowledge [8] now, yet in the case of
heterogeneous catalysis the knowledge of the active site geometry is of importance for academia
and industry. Another important issue of the catalyst performance is a catalyst deactivation, and then
mechanisms of deactivation have to be specified. This problem was not in the focus despite a
growing number of publications on the catalytic BDO conversion.
Concerted action of acid-base pairs (oxygen vacancies and oxygen anions) on the surface of
ZrO modified by alkaline earth oxides was proposed by Duan et al. in [9,10] for the conversion of
2
Page 2 of 39

BDO to 3B2OL. Probably the same mechanism is valid for highly active and selective Sc O
2
3
catalyst in the same process [11,12]
Over 90% of selectivity to BD at 99.9% conversion of BDO was reported by Tsukamoto et al.
13] using 10% CsH PO /SiO catalyst. On the basis of FTIR and SEM-EDX data the authors
[
2
4
2
proposed isolated silica grafted orthophosphate as an acid-base active pair having acidic P-OH
-
+
moiety and basic P-O ∙∙∙Cs moiety. At the same time unsupported CsH PO had the same
2
4
selectivity pattern in BDO dehydration, though it is less probable that calcined CsH PO has
2
4
isolated orthophosphates.
Though concerted action of acid-base pairs is important in the mechanism of BDO
dehydration to BD/3B2OL according to the research cited above, it may not be applicable to any
catalyst. Effects of sodium addition to alumina were thoroughly studied in [14]. Indeed, additional
basic sites were created, but this led to the decrease of the selectivity to BD.
Isolated basic centers of alkali-doped silica [15] play a key role in the dehydration of vic-diols
to epoxides. Cs/SiO was reported to be the most active catalyst for this process and the activity of
2
the catalysts increased in the order Na<K<Cs.
Taking into account the research papers and reviews on the dehydration of diols [3–6,15–17],
it is possible to conclude that a set of four parallel pathways (Scheme 1) describes the dehydration
process. The first route represents a traditional rearrangement mechanism via 1,2-hydride shift
leading to MEK or via 1,2-methyl shift leading to isobutyraldehyde (IBA). Since hydride shift is
faster than methyl shift, then MEK dominates over IBA. The second route is a sequential 1,2-
elimination of water leading to 3B2OL after the first dehydration and BD after the second one.
According to the mechanism of elimination reactions [8] MEK and IBA may be formed by e.g.,
standard acid-catalyzed E elimination, while 3B2OL and BD are formed by E -type elimination.
2
1cB
The third route represents dehydrative epoxidation leading to BTO, and S 2 mechanism via
N
concerted action of acid-base sites was proposed for this reaction [15]. The ratio of selectivities
(SBD+S3B2OL)/(SIBA+SMEK) may characterize the relative rates of the two major pathways of BDO
Page 3 of 39

dehydration. For the convenience, we will term it elimination/rearrangement ratio (E/R). The fourth
route represents dehydrogenation pathway taking into account detectable amounts of acetoin (Actn)
and butanedione (BDOn) found in the products. [6].
In the present work we would like to analyze the nature of the active centers in silica
supported alkali phosphate catalysts and interpret deactivation patterns during the conversion of
BDO.
2. Experimental
2
.1. Catalyst preparation
Davisil type silica grade 636, 36-60 mesh (Sigma-Aldrich) was used as a support. Before the
impregnation step silica gel was calcined at 500 °C for 3 h. Calculated volumes of sodium,
potassium hydroxide solution (both Samchun Chemicals), cesium carbonate (Sigma-Aldrich)
solution, and orthophosphoric acid (Sigma-Aldrich) solution, were mixed in the required
proportion. Deionized water was added to maintain the volume required for the incipient wetness
preparation.
A vacuum impregnation method according to Lu et al. [18] was employed. Before the
-2
impregnation silica gel was evacuated at <10 Torr at 350 °C for 3 h. After cooling down to an
ambient temperature under vacuo a solution of alkali phosphate was added to the gel and left for 1
h. Thus obtained wet powder was evacuated at 90 °C for 2-3 h, finally the dry powder was calcined
in air at 500 °C for 3 h (heating rate 2 °C/min). A list of catalysts and some textural characteristics
are presented in Table S1. Abbreviations for the samples used in this study follow that of our
previous publication [6] i.e., silica supported cesium phosphate having ratio Cs/P=1.75 (mol/mol)
and 20 wt.% of cesium phosphate phase (Cs O) P O is abbreviated as 1.75Cs
_
P/SiO2 20.
_
2
1.75
2
5
2
.2. Catalyst characterization
Specific surface area (SSA) and porosity of the samples were determined by N adsorption-
2
desorption isotherms at 77 K using ASAP 2020 or Tristar II apparatuses (Micromeritics). The
samples were evacuated at 350 °C overnight before the measurements.
Page 4 of 39

Acid-base site density and distribution were determined using the temperature programmed
desorption (TPD) of ammonia and carbon dioxide conducted by AutoChem II (Micromeritics).
About 0.4 g of the catalyst was pretreated in a helium flow (50 cm³/min, 550 °C, 1 h), cooled down
to 100 °C and exposed to 50 cm³/min flow of pure CO or 15% v/v NH /He flow to saturate base or
2
3
acid sites. Physically adsorbed species were removed by He flow (100 °C, 3 h) and then a TPD
trace was recorded at 10 °C/min heating rate and 50 cm³/min He flow. TPD cycles were repeated 2-
3
times to until good reproducibility. TPD traces were deconvoluted by Gaussian lines permitting
better quantification of low (LT), medium (MT), and high temperature (HT) desorption and the
number of corresponding sites was quantified based on the deconvolution results. In addition
desorbed ammonia was trapped by 0.01 mol/L solution of H SO and desorbed carbon dioxide was
2
4
trapped by 0.05 mol/L NaOH solution. The total amounts of desorbed species were checked by a
titration to eliminate possible effect of water vapors or other errors.
Diffractograms of finely powdered samples were collected on Xpert Pro (PANAlitical) using
CuKa radiation at 45 kV power and 40 mA current. Since some samples were highly hygroscopic,
the collection was performed both at 25 and 350 °C using the Anton Paar TCU 2000 heating stage.
Raman microprobe spectrometer Nanofinder Flex G (Tokyo Instruments, Japan) equipped
with a solid state laser in the backscattering configuration using the 10x objective lens was used to
collect Raman spectra of the catalysts. The spot size was about 2 m, and the laser power was kept
below 0.1 mW using 532 nm excitation wavelength.
Acid properties of supported catalysts were characterized using an FTIR spectroscopy of
adsorbed pyridine (Py). This study was conducted using a Nicolet Nexus 670 instrument equipped
2
with DTGS detector. The samples were pressed into self-supporting pellets (20-30 mg, 0.8 cm ),
placed into the IR cell, and evacuated at 300°C for 1 h. After cooling to 100°C, Py vapor (1 Torr)
was introduced, and spectra were taken at 30 and 60 min. Desorption of Py (by stepwise evacuation
for 30 min) was performed at 25, 50, 100, 150 and 200°C, and the corresponding spectra were
-1
registered (64 scans at 4 cm resolution).
Page 5 of 39

2
.3. Catalytic Activity
For short contact time experiments 0.1-0.22 g of catalyst mixed with 3 g of silica sand
(Samchun Chemicals) were packed between quartz wool plugs in the center of 50 cm long tubular
1
/4'' reactor. 0.5 mm glass beads filling all the void space of the reactor were used to suppress non-
specific reactions. For long contact time experiments 5-8 g portion of catalyst was charged into the
quartz tubular (3/8'') reactor filling all the void space.
Liquid BDO (97% Sigma-Aldrich) was fed by a syringe pump (0.1-0.25 cm³/h) into the empty
5
L volume kept at 180 °C to eliminate concentration fluctuations (caused by the syringe pump
delivery and droplet evaporation into the gas phase). Nitrogen carrier gas was used to maintain
desired partial pressure of BDO. All communications were maintained at 180 °C under the typical
reaction conditions. 1 atm system pressure was kept constant in all the experiments.
The reactor was held at a constant temperature in the 5-zone oven regulated by individual
thermocouples. Before each experiment the catalyst was pretreated in N flow (100 cm³/min) at
2
4
00°C for 1 h (heating rate 10°C/min).
The effluent gas was analyzed by an online GC (Agilent 6890N) using an FID detector and
Restek RTX-VRX capillary column (60 m). For identification purposes liquid products were
collected and analyzed by an offline GC/MSD (Agilent 5973) using the same capillary column.
In the typical catalytic run a catalyst was kept for 1 h at 400°C under the flow of nitrogen,
then BDO feed was introduced. The sampling was continued for at least 20 h to evaluate a
dynamics in the performance.
3. Results
3
3
.1. Characterization
.1.1. Textural properties
2
1
0% wt. silica supported alkali phosphates show a variation of SSA within 200-400 m /g
2
upon the change of M/P ratio, while 20% wt. samples have lower SSA values within 65-320 m /g
(Table S1). SSA values are dependent on the loading, nature of cation, and M/P ratio.
Page 6 of 39

At M/P=1~1.5 characteristic trend in SSA variation is observed (Fig. 1). Common maximum
2
2
of SSA (~400 m /g) at M/P=1 followed by a common drop of SSA (~300 m /g) at M/P=1.5 was
observed for all three sets of catalysts at 10% wt. loading (Fig. 1A). At 20% wt. loading the same
2
maximum of SSA (within 230-300 m /g) was observed for all sets of catalysts, followed by a
2
common drop of SSA values (90-230 m /g) at M/P=1.25 (Fig. 1B).
At 1.75≤M/P≤2.5 SSA values are dependent on the nature of cation, and a clear split into 3
distinct regions of relatively steady SSA may be noted (Fig. 1). Sodium-based samples show the
2
2
highest SSA values (close to 400 m /g at 10% wt. loading and 250-300 m /g at 20% wt. loading).
2
Potassium-based samples show intermediate SSA values (~300 m /g at 10% wt. loading and 150-
2
2
50 m /g at 20% wt. loading). Cesium-based samples show significantly lower SSA values (~250
2
2
m /g at 10% wt. loading and 80-160 m /g at 20% wt. Loading).
The reproducibility of SSA values was checked in a number of batches (duplicate and
triplicate preparation) at M/P=0.75-0.85 for Na-, K-, and Cs-based samples. Batch to batch
variation of SSA was less than 5% provided identical protocol based on vacuum impregnation.
3
.1.2. XRD
Samples loaded with 5-20% wt. of alkali phosphates did not reveal any XRD pattern except
the diffuse line typical to amorphous silica. For the present catalytic study 10-20% wt. loading was
used, and thus it was difficult to evaluate the phase composition directly. In order to get an indirect
account of the dominating phases, non-supported phosphates were prepared the same way but
omitting the addition of the silica support. Unsupported sodium-based samples revealed a good
crystallinity, while potassium- and cesium-based samples still had inferior crystallinity (Fig. 2).
1
.75Na
the latter is dominating). 1.75K
sample we could not find any good matches due to rather poor crystallinity of this sample. (Just
_
P sample clearly demonstrates the existence of pyrophosphate and tripolyphosphate phases
(
_
P sample shows only tripolyphosphate phase, while for 1.75Cs
_
P
from the stoichiometric point of view, a formation of tripolyphosphates and pyrophosphates for this
particular M/P ratio is expected.)
Page 7 of 39

3
.1.3. Raman spectroscopy
Raman spectroscopy may provide better details on the structure of phosphates motifs
stabilized on the surface of silica. First, it is important to note that supported alkali phosphates are
quite labile during thermal treatments. In the case of freshly calcined at 500°C supported samples
bands characteristic to phosphates disappear in the most instances and reappear after storage (Fig.
3
A). Absence of any bands in spectra after the calcination is indicative of highly disordered state of
phosphate layer on the surface of silica.
Na P/SiO2 20 sample demonstrates intense Raman band at 936 cm that disappears after the
calcination at 500°C (Fig. 3B). Immediately after the calcination a new band at 1027 cm develops.
-1
3
_
_
-1
-1
This effect implies the disappearance of orthophosphates (characteristic band at 936 cm ), and
-1
formation of pyrophosphates with a characteristic band at 1027 cm .
K- and Cs-based samples demonstrate weaker bands compared to Na-based catalysts (Fig.
3
C). Supported alkali phosphates at M/P=1.75 exhibit characteristic Raman bands that can be
attributed to pyrophosphates and tripolyphosphates. The most intense bands of supported alkali
-1
phosphates (with M/P=1.75) are located at 1011-1019 cm , very close to the corresponding band of
-1
pure sodium pyrophosphate [19]. Less intense bands — at 1093-1097 cm , are assigned to the
central PO symmetric stretching mode of P O chain [20,21] in tripolyphosphates.
2
3
10
Raman spectra of “acidic” samples (M/P=0.5-1) are presented in Fig. 3D. Two main bands at
-1
6
77 and 1167 cm of low sodium 0.5-1Na
_
P/SiO2 20 samples correspond to symmetric stretching
_
2
vibrations of P-O -P bridges and PO (Q ) species [22]. It means that low sodium supported
b
2
phosphate is composed mostly of chain or ring structures as expected from the chemistry of
inorganic phosphates [23]. Raman bands are better resolved in the case of 0.5Na P/SiO2 20 sample
than for 1Na P/SiO2 20 indicating better ordered structure for the former sample.
.1.3. TPD
Standard ammonia and carbon dioxide TPD was used for the characterization of density and
strength of acidic and basic sites. Pristine silica gel shows low density of acid sites while addition of
_
_
_
_
3
Page 8 of 39

alkali or phosphoric acid to silica support leads to more intense TPD traces (Figs. 4-8). Silica gel,
modified by phosphoric acid, provides the highest density of weak acidic sites according to the
desorption peak with a maximum around 200 °C (Fig. 4A). Addition of small amounts of alkali to
phosphoric acid (M/P=0.5) leads to the dramatic change in TPD. This is manifested by the
appearance of additional desorption peaks within 250-350 °C for K
_
P, Cs
_
P samples and 350-450
°
C for Na P sample (Fig. 4A). Therefore, stronger acid sites emerge upon the introduction of alkali
_
to supported phosphoric acid that seems to be a paradox from the text-book point of view.
Moreover, the same effect is observed in carbon dioxide TPD: addition of even small amounts
of alkali immediately leads to the appearance of desorption peaks within 300-500 °C corresponding
to HT region (Fig. 4B). Therefore, these new CO adsoption sites are stronger than silanolates Si-O-
2
Na shown for comparison in Fig. 4B. NaOH modified silica exhibits only LT desorption.
In parallel with the results obtained by Raman spectroscopy, TPD study revealed that
supported alkali phosphates are quite labile during thermal treatments. Several consecutive TPD
cycles show substantial change of TPD curves (Fig. 5): during consecutive heating and cooling
cycles the amount of desorbed ammonia at LT range increases and at HT range — decreases. Six
month storage of the catalyst under the ambient conditions leads to ca. 1.5-fold increase of the peak
area within 250-500 °C (Fig. 5B). Obviously these changes imply an increase of the density of
weaker sites and a decrease of the density of stronger sites upon the progress of cyclic thermal
treatments. Long-term storage under ambient conditions leads to the pronounced increase of the
density of stronger acid sites.
Comparable changes of TPD traces are observed upon the decrease of alkali phosphate
loading of the catalyst (Fig. 6). For silica loaded by lower amount of alkali phosphates (5-10% wt.)
more ammonia is desorbed at the lower temperature range and less — at higher temperature. Low
temperature desorption in this case may be a single peak (Fig. 6B) or a peak with a composite
structure and a maximum around 200 °C (Fig. 6A). For silica loaded by a higher amount of alkali
phosphates (~20% wt.) desorption at HT (~300 °C) and LT (~200 °C) are better resolved (Fig. 6B).
Page 9 of 39

It is important to take a note that a TPD curve of 2% wt. Na O/SiO sample was almost identical to
2
2
that of 5% wt. Na P/SiO sample (Fig. 6B). In general, a change in the loading causes a
_
2
corresponding change in dispersion/crystallinity of supported phase. Then the regularities in the
change of the shape of TPD traces with loading may be related to the different dispersion of alkali
phosphate on the surface of silica.
An example of ammonia TPD curves for sodium-, potassium-, and cesium-phosphates at
M/P=2 is presented in Fig. 7A. TPD traces are strongly dependent on the nature of cation. Addition
of any alkali phosphate to pure silica leads to the increased amount of desorbed ammonia compared
to pure silica support. Peak at 150 °C and broad desorption within 250-400 °C are observed for a
sodium-based sample. Considerably more intense desorption traces are registered for potassium-
and cesium-based samples indicating higher density of acid sites. Two broad desorption features are
clearly observed at LT-MT (150-300 °C) and HT (400 °C) regions. We tentatively attribute LT
desorption within ~100-200 °C to physically bound ammonia, since LT desorption depends on the
time of the preliminary nitrogen purge at 100 °C. Desorption peaks within MT range ~200-300 °C
may be attributed to weakly adsorbed ammonia, while desorption peaks within HT range ~300-
5
00°C may be attributed to the centers of medium strength. This is just an arbitrary notation.
According to all accumulated TPD data there is a considerable difference within sodium-
based samples on one hand, and potassium-, cesium-based samples on another. Provided the same
M∕P ratio, sodium-based catalysts demonstrate lower density of medium acid centers, while
potassium and cesium-based catalysts have considerably higher density of medium acid sites.
An example of carbon dioxide TPD curves for sodium-, potassium-, and cesium-containing
samples and silica support is presented in Fig. 7B. In the same manner as for ammonia TPD, CO2
TPD traces are strongly dependent on the nature of cation. Addition of any alkali phosphate to pure
silica leads to the increased amount of desorbed carbon dioxide in TPD traces compared to silica
support. Broad weaker peaks at ~100-200 °C and ~250-350 °C are registered for a sodium-based
sample. Much more intense desorption traces are observed for potassium- and cesium-based
Page 10 of 39

samples. CO TPD curves of K- and Cs-based samples have two broad regions with maxima at 150-
2
2
80 and 300-400 °C. LT desorption region at 100-200 °C was attributed to physically bound carbon
dioxide, while MT peaks at 200-300 °C − to weak basic centers. HT peaks at 300-500°C − to the
centers of medium basicity. A notable difference between the sodium-based samples on one hand,
and potassium-, cesium-based samples on another is evident. In general, according to all the
accumulated data sodium-based catalysts demonstrate lower density of medium basic centers, while
potassium and cesium-based catalysts have higher density of medium basic sites.
It is important to note a certain qualitative, or even semiquantitative similarity between TPD
traces of ammonia and carbon dioxide displayed in Fig. 7. It seems like both density and strength of
acid and basic sites are related to each other. This effect suggests probable conjugation of the acid
and basic sites.
Fig. 8 shows an example of M/P ratio effect on the shape of TPD curves. Certain regularity
may be noted comparing TPD curves for “acidic” sample (Cs/P=1), “neutral” sample (Cs/P=2), and
“
basic” sample (Cs/P=2.5). More “acidic” 1Cs
and medium acid sites according to TPD curve (Fig. 8A): strong peak at 150-200 °C corresponds to
very weak (physical) adsorption of ammonia. “Neutral” 2Cs P/SiO2 20 sample clearly manifests
the highest density of weak and medium acid sites according to the most intense desorption curve of
ammonia within 200-450 °C. For more “basic” 2.5Cs P/SiO2 20 the density of weak and medium
_
P/SiO2
_
20 sample in reality has low density of weak
_
_
_
_
acidic sites decreases. Therefore, according to TPD traces the samples with M/P~2 have a topmost
density of weak and medium acid sites just by direct qualitative comparison.
The same trend is observed for carbon dioxide TPD curves (Fig. 8B). The most intense TPD
trace within 300-500 °C is registered for “neutral” 2Cs
.5Cs P/SiO2 20 shows ca. 50% smaller amount of desorbed CO . Low basicity level is observed
for 1Cs P/SiO2
_
P/SiO2
_
20, while more “basic”
2
_
_
2
_
_
20. Therefore, the highest density of weak and medium acid-base sites may be
assigned to the same “neutral” 2Cs
of acid and basic sites.
_
P/SiO2 20 sample. This observation suggests possible relation
_
Page 11 of 39

Accurate quantification of the acid and basic sites is an intricate task assuming the complex
structure of TPD curves, multiple peaks, tailing etc. But a real headache is the fact of considerable
change of the curve shapes during consecutive TPD cycles or dependence of the curve shape on the
catalyst history (Fig. 5).
Quantification of acid and basic sites may provide considerable scattering of data points
depending on the history of the catalyst. For example, 20% wt. sodium-based samples did not show
any specific dependence of the total desorption of ammonia on Na/P ratio (Fig. S1) due to the
considerable scattering of data depending on the batch of a catalyst and shelf storage.
An effect of M/P ratio on the density of acid-base sites was better resolved for K- and Cs-
based samples. Evident maximum of the density of both medium acid and basic sites is observed
within Cs/P=1.5-2.5 (Fig. 9). Samples with lower loading (10% wt.) of cesium phosphates show
st
d
higher magnitude of deviation from the 1 cycle to 3 TPD cycle (Fig. 9A-B) compared to higher
loading (20% wt.) of phosphates (Fig. 9C-D). Hence, catalysts with low loading of phosphate are
more susceptible to the effects of a sample biography.
The observed trends in acid and base site density are not related to the trends of SSA change
with M/P ratio. Indeed, direct comparison of Fig. 1 and Fig. 9 for Cs-based set of catalysts cannot
provide any plausible grounds for the relation of textural effects and acid-base properties.
3
.1.4. FTIR of adsorbed pyridine
The nature of acid sites was characterized using a well documented transmission FTIR
spectroscopy of adsorbed Py [24]. After adsorption of Py vapors at 100°C all alkali-phosphate
-1
-1
samples, as well as pure silica support, exhibited IR bands at 1444 cm and 1595 cm . The latter
-1
band has a shoulder at 1580 cm (Fig. 10). These sites are indeed weakly acidic according to a
relatively low temperature required for the complete removal of adsorbed probe molecules. The
disappearance of all Py-related IR bands takes place upon the evacuation at 150°C (Fig. S2). Pure
silica support retains more Py compared to the supported alkali phosphates at any desorption step.
3.2. Catalytic Activity
Page 12 of 39

Conversion of BDO by phosphoric acid supported on silica leads to the products of
rearrangement (Fig 11B). P/SiO2 10 catalyst provides complete conversion over 30 h time on
_
stream (TOS) – Fig. 11A and very stable selectivities to major products (Fig. 11B). Addition of even
small amounts of alkali metals (M/P=0.5) leads to the impressive changes in the catalyst
performance. Initially all 3 set of 0.5M
certain TOS (the latter is dependent on the nature of cation) catalyst deactivates. Deactivation starts
after ~20 h of TOS for 0.5Na P/SiO2 10 sample, ~10 h for 0.5K P sample, and ~5 h for 0.5Cs
_
P samples shows complete conversion of BDO but after
_
_
_
_
P
sample. The nature of alkali cation affects the deactivation rate: Na<K<Cs (Fig. 11A). As soon as
deactivation starts, the change of the selectivity pattern occurs simultaneously. For deactivated
samples elimination becomes more favorable according to the increase of E/R ratio (Fig. 11B-C).
The process of deactivation proceeds for a few hours until the next steady state is achieved with a
more stable conversion and a selectivity pattern: E/R=1.4 for Na-, E/R=2.8 for K- and E/R=4.5 for
Cs-based samples. All these changes are indicative of a primary deactivation of the more active acid
sites responsible for rearrangement and relative resistance the sites responsible for elimination.
During the deactivation of 0.5M P/SiO samples characteristic changes of BTO selectivity are
_
2
observed (Fig. 11C). As soon as the catalysts lost their initial activity towards rearrangement, not
only E/R ratio increased, but simultaneously selectivity to BTO increased considerably. This effect
may imply that the sites responsible for elimination are also active in the dehydrative epoxidation
route.
In general at any M/P ratio and phosphate loading sodium-based catalyst show much more
stable conversion compared to K- and Cs-based catalysts (Figs. 11A, 12A). While
1
.75Na
_
P/SiO2
_
20 sample demonstrated stable 80% conversion in the conditions of short contact
20 and 1.75Cs P/SiO2 20 catalysts under the same conditions show fast decay
time, 1.75K
_
P/SiO2
_
_
_
of conversion from 70% to 50-30%. At the same time during the catalytic run an increase of E/R
ratio is observed for these samples (Fig. 12B) suggesting deactivation of acid sites responsible for
the rearrangement.
Page 13 of 39

Taking into account stepwise dehydration of BDO → 3B2OL → BD it was logical to check
the catalytic performance of supported alkali phosphates in the conversion of 3B2OL. In this case
BD was a major product of the reaction with a smaller amounts of crotyl alcohol (CA) and MEK
(
Table S2). The latter products obviously appeared due to the isomerization of 3B2OL. What is
more important to note — it is a faster deactivation of the catalysts during the conversion of 3B2OL
Fig. 13) compared to BDO conversion (Fig. 12A). Thus 3B2OL plays more important role in the
(
deactivation than BDO.
It is necessary to mention as well that all the studied catalysts generate products of BDO
dehydrogenation (butanedione – BDOn and acetoin – Actn), at selectivities within 0.5-20% (Table
1
). Both acidic (M/P<1) and basic (M/P)>2 catalysts provided higher levels of dehydrogenation.
Temperature dependence of BDO conversion may provide additional information on the
mechanism of the process. It is evident that the catalyst should not be deactivated during the tests.
Strictly speaking this is impossible according to Figs. 12-13. K- and Cs-based catalysts start to
deactivate after 5 h TOS during BDO conversion and almost immediately during 3B2OL
conversion. To reduce the effects of deactivation partial pressure of BDO in the feed may be
lowered. At 0.1 ml/h BDO good reproducibility of catalytic data (within 10% of variation) at
different temperature profiles was obtained (Fig. S3). Arrhenius plots for M/P=1.75 are presented in
Fig. 14. Corresponding apparent activation energies are presented in Table 2. Lower experimental
activation energy for dehydrative epoxidation is evident for any supported alkali phosphate.
Considerably lower activation energy for elimination and rearrangement is observed for Cs-based
catalyst.
4. Discussion
4
.1. Textural properties
i
Alkali polyphosphates may be viewed as a combination of PO tetrahedra using accepted Q
4
notation, where i is the number of bridging oxygens per tetrahedron [23]. Upon the addition of
alkali metal oxide M O to phosphorus pentoxide P O , and the increase of the M O:P O ratio
2
2
5
2
2
5
Page 14 of 39

0
(
M/P), a three-dimensional network of P O is progressively depolymerized, and the numbers of Q ,
2 5
1
2
Q and Q tetrahedra increase. These changes lead to the formation of chains, rings and finally
isolated PO tetrahedra. When alkali phosphates are supported on silica, additional effects of the
4
interaction with a support may change the composition of phosphate polymers.
Characteristics of textural properties presented in Table S1 and Fig. 1 indicate considerable
variation of SSA with M/P ratio and loading, but a phenomenon of a single maximum of SSA at
31
M/P=1 for all three sets of catalysts requires certain interpretation. According to P NMR study of
1
2
3
alkali phosphate glasses, at M/P=1 there is a point of singularity with respect to Q , Q and Q site
2
fractions [25,26]. M/P=1 corresponds to [Q ]fraction=1 that may lead to a certain optimum of
polyphosphate chain packing on the surface of silica. Since this effect appears only at M/P=1 then
alkali metaphosphates have such a packing on the surface of silica that resembles a monolayer
coverage without pore blocking and/or changing the local geometry of the surface. This means a
better interaction of alkali phosphate with silica at M/P=1. Schematically this effect is illustrated in
Fig. 15. (KPO ) chain (potassium was used here just as an example of an alkali cation) is
3
n
connected to the surface via electrostatic interaction of potassium cations with surface silanols and
bridging oxygens of silica (Fig. 15C).
In this case it is necessary to consider the distance between adjacent silicon atoms on one
hand, and the distance between adjacent PO tetrahedra in metaphosphate chain on another, in order
4
to find out if a fit between two polymers is possible. Silica gel is not a well-ordered structure and
there is a certain variation of Si-Si distance. Moreover, due to the dehydroxilation at conditions of
catalysis Si-O-Si bond angle in newly formed siloxane bridges may change leading to a propagated
change of the whole silica network [27]. At 400 °C Si-O-Si bond angle of the newly formed
siloxane bridges may be around 105° [27] that leads to Si-Si distance ~2.6 Å (assuming Si-O bond
length ~1.61 Å). Si-Si distance then may vary from 2.8 to 3.1 Å assuming that inter-tetrahedral Si-
O-Si bond angles are within 120-150° and Si-O bond length is about 1.61 Å in average. Si-O bond
length in various modifications of silica is within 1.54-1.69 Å and inter-tetrahedral Si-O-Si bond
Page 15 of 39

angle is varied from 120 to 180° [28]. Tabulated P-O distances and P-O -P angles in
b
b
metaphosphate glasses may be found in [23,29,30], and according to these works P-O distance is
b
varied between 1.56 and 1.61 Å, while P-O -P angle — from 129 to 142°. Thus, P-P distance may
b
vary from 2.8 to 3.0 Å that is in a good match with Si-Si distances in silica. Therefore, interaction
between silica and polyphosphates as displayed in Fig. 15 is possible due to the good fit and certain
flexibility of P-O -P angle, as well as relatively free positioning of alkali cation around P-O bond
b
[31].
For M/P<1 (ultraphosphates) considerably lower SSA values are observed. Ultraphosphates
are crosslinked polyphosphates according to the stoichiometry. In the point of crosslink PO4
tetrahedron have to be rotated in a way to accommodate a proper positioning of the polyphosphate
branch on the surface of silica (Fig. 15D). This may lead to the formation of folded structures, loops
and other “imperfections” leading to the less compact accommodation of polyphosphates on the
surface.
Decline of SSA at M/P~1.25-1.5 is less obvious for the clarification. First, it is less probable
that only a linear polymer is stabilized on the surface at M/P=1. It is more probable that the
distribution of crosslinked chains, rings, linear polymers, and some residual grafted phosphoric acid
are present on the surface of silica (Fig. 16A) as well, since phosphoric acid interacts with silica via
formation of Si-O-P bonds [32–35]. Upon the addition of M O isolated grafted phosphates are
2
unstable [33] and transformed into phosphate polymers causing a decrease of SSA if the number of
crosslinks is increasing. According to Georgelin et al. [36] silica has certain stabilizing effect on the
dehydration of supported KH PO . Thus, shorter phosphate polymers tend to form on the surface of
2
4
silica compared to the bulk phosphate phase. The same effect may be valid for the M O addition to
2
silica-stabilized polyphosphate since a part of M O may interact with silanols.
2
Upon the addition of alkali (M/P>1.5) further depolymerization of polyphosphates to short
species like tripolyphosphates, pyrophosphates and orthophosphates takes place leading to higher
values of SSA. Sodium-based samples at M/P~1.75-2.5 show the highest SSA values, potassium-
Page 16 of 39

based samples — intermediate values of SSA, and the lowest values are observed for cesium-based
catalysts. Taking into account the results of Raman study (Fig. 3C) Na P series show the most
intense Raman bands compared to K P and Cs P series. Our XRD data (Fig. 2) also indicate that
_
_
_
sodium phosphate has much better crystallinity compared to potassium and especially cesium
phosphates. Therefore, XRD and Raman data reflect more regularly organized packing of sodium
phosphate phase and thus Na P samples may be in a more compact state of phosphates.
_
The size of alkali cations plays an important role in the packing of phosphate phase due the
regular decrease of SSA in the row Na>K>Cs at the same loading of phosphate. Indeed, visual
comparison of geometric parameters (Fig. 15B) suggests that K- and Cs- based polyphosphates may
have some difficulties in packing on the surface of silica due to the big size of cations at higher M/P
+
ratios. The size of Cs exceeds Si-Si distance in silica, that obviously may lead to the problems of
phosphate packing on the surface of silica support.
Obvious drop of SSA for basic phosphates at M/P=2.5-3 may be related to alkali induced
shrinkage of silica [37].
4
.2. Interaction of alkali phosphates and silica
Thermal treatment induces a dramatic disorder in the structure of alkali phosphates stabilized
on the surface of silica according to the results of Raman and TPD studies. Calcination at 500°C
leads to featureless Raman spectra (Fig. 3A) indicating induced heterogeneity in a local structure of
PO units leading to the broadening of characteristic vibrational bands beyond detection.
4
TPD study of the same samples also revealed high sensitivity to thermal treatment. TPD
traces of ammonia support the idea that the dispersion of alkali phosphate increases after several
thermal treatments. First, samples with low loading of alkali phosphates (and thus better dispersion)
have intense TPD peaks in LT region according to Fig. 6A. Second, 1.9% wt. Na O/SiO and 5%
2
2
wt. 1.82Na P/SiO show almost identical TPD traces (Fig. 6B). The latter may imply a migration of
_
2
sodium cations of sodium phosphates to acidic silanols of support. The process of the migration of
alkali cations to silanols is supported by Raman spectra after the thermal treatment of
Page 17 of 39

0
3
Na_
P/SiO2 20 sample (Fig. 3B). In the latter case Q units of orthophosphates manifested by 936
_
-1
1
-1
cm band are replaced by Q units of pyrophosphates manifested by the band at1027 cm .
Migration of sodium cations to acidic silanols of silica, followed by the dimerization of
hydrogen phosphate units may explain changes in Raman spectra of silica supported sodium
orthophosphate:
+
-
Na PO + H O ↔ Na + Na HPO + OH
(1)
(2)
(3)
(4)
3
4
2
2
4
-
-
SiOH + OH ↔ SiO + H O
2
-
+
SiO + Na ↔ SiONa
2
Na HPO → Na P O + H O↑
2
4
4
2
7
2
Evidently reactions (1-2) may proceed at the ambient conditions. Reaction (3) may be
responsible for the similarity of ammonia TPD curves of 1.9% wt Na O/SiO and 5% wt
2
2
1
.82Na
_
P/SiO samples. Reaction of phosphate dimerization (4) requires thermal treatment.
2
Disappearance of Raman bands attributed to pyrophosphates and tripolyphosphates after
thermal treatment (Fig. 3A) may be attributed to the same processes of hydrolysis-migration-
oligomerization. In this case reactions (1,4) have to be modified in the correspondence with the
structure of e.g., pyrophosphate:
+
-
Na P O + H O ↔ Na + Na HP O + OH
(1')
4
2
7
2
3
2
7
2
Na HP O → Na P O + H O↑
(4')
3
2
7
6
4
13
2
Ideally corresponding Raman bands tetrapolyphosphate should be observed, but quite often
Raman bands of polymeric phosphates supported on silica are not observable or quite weak due to
the interaction of phosphate chains with the support. Due to this interaction multiple deviations of
local P-O-P symmetry occur leading to the broadening of vibrational bands beyond detection.
4
.3. The nature of acid and basic sites
According to the standard chemistry of alkali phosphates and silica the following candidates
+
+
+
for acid and basic sites may be considered: (1) Lewis acid sites — alkali cations Na , K , and Cs ;
(2) Brønsted acid sites — silanol groups in silica ≡SiOH and hydroxyl groups of phosphates -P-OH;
Page 18 of 39

-
(
3) base sites — -P-O anions of phosphates. Interaction between alkali phosphates and silica
-
support may create additional sites, like silanolates ≡SiO by a cation transfer according to the
reactions (1-4, 1', 4') above. Chemical interaction of NaH PO with silica is improbable while
2
4
H PO is able to interact with silanols according to DFT simulation [33]. Phosphoric acid grafted to
3
4
silica surface shows a formation of Si-O-P linkages [32,34,35,38]. P/SiO2
C shows mild acidity according to ammonia TPD with a single peak at 200 °C (Fig. 4). This
corresponds to the mild acidity detected by Kawi et al. for H PO /Si-MCM-41 calcined at 400 °C
_
10 sample calcined at 500
°
3
4
[
35]. Usually these mild acid centers of H PO /SiO are attributed to isolated grafted species of
3 4 2
Brønsted type like (Si-O) -P=O(OH) [32] (Fig. 16A). Polymerization of H PO during the
2
3
4
dehydration of silica is less probable than chemical interaction with silanols according to DFT
simulation [33]. Then it is logical to assume that these grafted units of phosphoric acid represent
acid sites responsible for the rearrangement and thus low E/R ratio according to catalytic data for
P/SiO2 10 sample (Fig. 11B).
_
As soon as alkali ions enter the scene even in small amounts (M/P=0.5) then quite specific
features both in TPD and catalytic dehydration of BDO are observed. Additional desorption peaks
of ammonia at HT region appear (Fig. 4). Alkali cations induce faster deactivation of the catalyst
(Fig. 11), and when deactivated the catalysts show stable conversion and much higher E/R (hence
the selectivity pattern changes from rearrangement to elimination). These two features imply that at
M/P=0.5 acid sites are present on the surface of fresh catalysts that are responsible for the
rearrangement. These sites provide LT peaks in ammonia TPD and low E/R ratio. They are much
more active than the sites responsible for elimination. The latter sites are revealed during the
catalytic run only when rearrangement sites are deactivated.
Unlike pure phosphoric acid, NaH PO on the surface of silica has too high barrier to react
2
4
with silanols and condensation of phosphate units is more probable according to DFT simulation
33]. Thus instead of expected partially neutralized grafted PO units (Fig. 16C “unfavorable
[
4
structure”) chains of metaphosphate are formed upon the introduction of alkali (Fig. 16B-C). Then
Page 19 of 39

the surface the samples at M/P=0.5 and M/P=1 is comprised of polyphosphates that provide
elimination (see lower) and isolated grafted phosphoric acid that provide rearrangement (Fig. 16C).
At the same time catalytic data show an increase of E/R ratio during the catalytic run (Fig. 12B)
suggesting that some isolated Brønsted sites exist at M/P>1 and slowly deactivate. Since at this M/P
ratio a formation of isolated phosphates is not favorable [33] then these centers may be viewed as
+
+
an acidic -PO(OH) end group due to H — M exchange by migration (Fig. 16B).
2
Additional experimental evidence in favor of linear polyphosphates being active centers in
catalytic dehydration of BDO vs grafted isolated species (Fig. 16C) may be revealed upon the
comparison of Raman, XRD, TPD, and catalytic data. Fig. 3D gives direct evidence that
metaphosphates are present in supported 0.5Na
_
P/SiO and 1Na P/SiO samples according to
_
2
2
-1
characteristic bands at 677 and 1167 cm . For the low loading of phosphates (5-10% wt) mostly LT
desorption peaks are observed, while for 20% wt. samples more intense HT desorption is observed
(Fig. 6). Low loading of phosphates should provide better opportunity for higher dispersion and
thus more isolated phosphate species that give rise to LT bands. At the same time deactivated
catalysts with M/P=0.5 (with deactivated isolated acidic phosphates) show higher E/R ratio close to
catalysts with M/P=1.5-2. The latter catalysts have mostly short pyrophosphate and
tripolyphosphate units (Fig. 2) and thus structurally they are closer to linear polyphosphates. It is
very highly likely that metaphosphate chains are continuing catalytic conversion after the
deactivation of isolated units.
According to the stoichiometry of ultraphosphates and metaphosphates (M/P≤1),
polymerization of phosphate units during the dehydration should lead to linear and crosslinked
polymers of infinite length or cyclic polymers that have only P=O , POM and P-O-P bonds. And
t
thus any Brønsted acidity should be impossible. In reality completely dehydrated ultra- and
metaphosphates cannot be obtained under practical conditions. Depending on a water vapor
pressure and annealing time polyphosphate chains of variable length are formed [39]. For example,
Graham's salt (NaPO ) in practice is a mixture of polyphosphates with an average length of 30-50
3
6
Page 20 of 39

P atoms [40]. Moreover, silica support reduces the extent of dehydration of KH PO /SiO [36] and
2
4
2
thus shorter chains of polyphosphate are expected to form. If shorter chains are formed then the
number of acidic P-OH end groups increases.
If HT desorption of ammonia is related to the interaction of ammonia with these P-OH end
groups of polyphosphate chains then it is possible to evaluate the average length of polyphosphate
chains assuming 1:1 stoichiometry of ammonia interaction with each end group. This calculation is
presented in Table S3. For example, 0.5Na
_
P/SiO2
_
10 sample has 25 P atoms length of polymer,
P/SiO2 10 sample has 16 P atoms length.
0
.5K P/SiO2 10 sample has 8 P atoms length, and 0.5Cs
_
_
_
_
These numbers are in line with practical expectations assuming more hygroscopic properties of K-,
and Cs-phosphates.
Then it is possible to interpret the TPD data for all M/P ratios. For M/P~1.5-2.5 acid centers
with medium strength are provided by the end groups of pyro- and tripolyphosphates. According to
3
1
Raman and XRD these species are dominating, earlier the same conclusion was obtained from P
MAS-NMR [6]. Acidic end groups may be formed by migration of cations to acidic silanols in the
same manner as it was shown earlier in reactions (1-3):
+
-
K P O + H O ↔ K + K HP O + OH
(5)
(6)
4
2
7
2
3
2
7
+
-
K P O + H O ↔ K + K HP O + OH
5
3
10
2
4
3
10
Taking into account that the affinity for silanols follows the row Li<Na<K<Cs [41], then
higher values of medium acidity detected for K- and Cs-based samples (Fig. 7) may be explained by
the difference in the affinity of alkali cations towards silanols.
Effects of the shelf life and thermal treatment onto the TPD curves may be attributed to the
partial hydrolysis of P-O -P bonds during the storage at ambient conditions with the formation
b
additional acid-base pairs that disappear during consecutive TPD cycles. Earlier Georgelin et al.
[
36] noted inhibiting effect of silica on the dehydration of supported KH PO /SiO . Due to the
2 4 2
relatively good electrostatic interaction of alkali phosphates with silanols polyphosphate chains tend
to be immobilized on the surface and thus cannot be dehydrated during one TPD cycle owing to the
Page 21 of 39

limited mobility and low probability of end-group collision. Then several TPD cycles are necessary
for the complete removal of additional P-OH end groups in order to get reproducible results (Fig.
5
). Then a simultaneous increase of weak acid sites upon thermal treatment is probably related to
the cation migration to silica with a formation of silanolates as in reactions (2-3). Silanolates
represent weak acid centers according to the experimental results for e.g.,1 .9% wt. Na O/SiO
2
2
sample (Fig. 6B).
Py desorbtion from the acid centers studied by FTIR provides an evidence of only weak acid
silanols that are coordinated by Py. The position of the two bands in Fig. 10, their similar intensity,
-1
single 1444 cm band imply that only hydrogen-bonded Py species are present. These species are
typical to Py adsorption on pure silica [42]. Therefore, FTIR-detectable Py is localized on silanol
groups of silica support in all silica-supported alkali phosphates and no Py species detectable by
FTIR were localized on P-OH Brønsted sites. The same phenomenon was observed earlier for
NaH PO -modified NaY zeolite [43]. While ammonia TPD showed clearly weak and medium acid
2
4
sites with increasing loading of NaH PO , adsorbed Py developed only IR bands typical to pure
2
4
NaY whatever was the loading of sodium phosphate. No bands of P-OH∙∙∙Py adducts were present.
This obvious contradiction between ammonia TPD and Py FTIR TPD may be explained by the
heterogeneity of P-OH local symmetry in supported catalyst causing line broadening beyond
detection.
HT desorption peaks of carbon dioxide detected for alkali phosphates is less easy to explain.
Silica doped by alkali metal show only LT peaks of carbon dioxide desorption (Fig. 4B) typical to
silanolates and attributed to mono- and bidentate carbonates [44]. Addition of even small amounts
of alkali metals to phosphates immediately leads to the appearance of stronger basic sites according
to desorption peaks within 300-500 °C (Fig. 4B). This means that these sites provide multidentate
adsorption of CO , like tridentate species in MgO [45]. Basic oxygen species in phosphates and
2
silanolates are the only plausible candidates for basic sites. Silanolates are ruled out since they
-
+
provide only LT desorption of carbon dioxide. Basic oxygen species of phosphates like PO ∙∙∙K are
Page 22 of 39

probable, but assuming that the concentration of P-O-M entities is much higher that of P-O-H end
groups then much more intense HT CO desorption should occur compared to NH HT desorption.
2
3
In reality both ammonia and carbon dioxide HT desorptions are comparable. We speculate that HT
-
+
desorption of carbon dioxide is provided by CO molecules bound to the end -P(OH)O ∙∙∙M groups
2
of linear polyphosphates (Fig. 16C). This gives a qualitative explanation why NH and CO HT
3
2
desorptions are so similar (Fig. 7). At the same time on the basis of a simple geometry interaction of
CO may proceed via tridentate complex: carbon interacts with negatively charged oxygen of
2
+
phosphate, one oxygen of CO interacts with cation M , another — with acidic proton of POH end
2
group (Fig. 16C). Such a tridentate coordination may explain the TPD peaks at 300-500 °C that are
absent in P O /SiO and M O/SiO samples.
2
5
2
2
2
Indirect support for the hypothesis that stronger adsorption of carbon dioxide is related to
basic oxygen of alkali phosphate conjugated to a nearby acidic POH group may be found in [46]. In
the latter study much more carbon dioxide desorbed at higher temperatures in the case of thermally
untreated potassium orthophosphate K PO . Increase of thermal pretreatment from 300 to 600 °C
3
4
decreased the amount of desorbed CO . Since K PO is hygroscopic and hydration will induce
2
3
3
hydrolysis with the formation of POH groups, then the latter groups are involved into the
interaction with CO . Anyhow at present a hypothesis about tridentate coordination of CO with
2
2
-
+
adjacent POH and PO M species (Fig. 16C) is speculative and additional evidence is required
(spectral data and/or DFT simulation). Anyhow, this hypothesis can nicely explain a conjugation
between basic and acid sites, experimentally observed in the present work in a convincing number
of experiments.
Earlier reported model of conjugated acid-base pairs in CsH PO /SiO catalysts by Tsukamoto
2
4
2
-
+
et al. [13] was based on the isolated grafted phosphate species with P-OH and P-O ∙∙∙Cs moieties.
Isolated acid-base pairs were proposed due to the absence of characteristic vibrations of P-O -P
b
bridging units in silica supported sample. In the same work [13] Tsukamoto et al. along with silica-
supported phosphates tested unsupported calcined CsH PO that (i) showed the same selectivity
2
4
Page 23 of 39

pattern in BDO dehydration as supported phosphate (ii) developed clear P-O -P bands in FTIR
b
spectra according to the same authors. Since calcined CsH PO is composed of chains and rings of
2
4
metaphosphates (CsPO ) but not isolated orthophosphates, then it is evident that polyphosphate
3
n
species should be also active in the same type of elimination reaction according to the complete set
of catalytic data by Tsukamoto et. al.
In the present work vibrations of P-O -P bridging units of NaH PO /SiO sample were
b
2
4
2
detected (Fig. 3D), thus metaphosphates stabilized on silica are present. Second, characteristic
changes of textural properties with M/P ratio are difficult to explain using only packing of isolated
phosphates species. Some regularities in the variations of SSA are better understood when
polymerization of phosphate species occurs. Moreover, DFT simulation performed earlier [33] did
not favor existence of isolated grafted partially neutralized orthophosphates. Instead, polymerization
of phosphate on the surface of silica is more favored. Assuming the flexibility of P-O -P bond
b
angles [23,29,30] and amorphous structure of silica, in the case of strong interaction with silica
surface phosphate bridges may have considerable variation of local symmetry causing broadening
of vibration bands beyond detection. Therefore, it may not be easy to detect FTIR bands P-O -P in
b
calcined CsH PO /SiO catalysts.
2
4
2
Doped ZrO and Sc O are very active in the conversion of BDO by elimination and are
2
2
3
comprised of acid-base pairs according to Duan et al. [9–12], that is in line with the current
research.
4
.4. Catalytic properties
Pure silica support has a low activity corresponding to acid-catalyzed BDO dehydration
(mostly rearrangement, E/R=0.6, poor mass balance ~45%) that was presented earlier [6]. Silica
doped by phosphoric acid display mostly rearrangement according E/R=0.4 (Fig. 11B). Even small
quantities of alkali metal added to the phosphate containing catalyst change both catalyst stability
and selectivity pattern. After deactivation, the catalysts provide stable conversion with quite
different selectivity pattern showing much higher E/R ratios up to 4.5 compared to initial E/R~0.5-1
Page 24 of 39

(Fig. 11). These changes of the selectivity may be interpreted in the following way: initially
0
.5M P/SiO catalysts have both isolated grafted phosphates that are Brønsted acids (Fig. 16A) and
_
2
conjugated acid-base pairs of polyphosphates (Fig. 16B). Isolated grafted phosphates catalyze
rearrangement and are much more active. At the same time these centers undergo a faster
deactivation, and when isolated acid centers are deactivated — conjugated acid-base pairs continue
conversion with much higher E/R ratio. The nature of alkali metal is important during deactivation,
since TOS before deactivation is 20h for Na-, 10 h for K-, and 5 h for Cs-based sample, while molar
amount of P is 1.2 mmol/g for Na-, 1.1 mmol/g for K-, and 0.7 mmol/g for Cs-based sample.
At M/P>1 the amount of isolated Brønsted centers is small since E/R>1.5 but may be not
negligible, because during the catalytic run gradual increase of E/R is observed (Fig. 12B). This
increase of E/R may be attributed to gradual deactivation/removal of isolated Brønsted centers, but
at M/P>1.5 existence of the centers like grafted phosphates (Fig. 16A) is questionable. Another
candidate for such center is -PO(OH) end groups of polyphosphates (Fig. 16B), but at present we
2
have no sufficient evidence about it, and additional experimental/simulation work is required to
verify this hypothesis.
E/R changes during the deactivation of 0.5M P/SiO catalysts are accompanied by a
_
2
simultaneous change in selectivity to BTO (Fig. 11). Thus there is quite a high chance that not only
silanolates [15], but basic oxygen in the end phosphate groups (Fig. 16C) may be responsible for
the dehydrative epoxidation route.
The levels of BTO increase with the increase of M/P ratio, and the size of cation is very
important (Table 1, Fig. 17A) in agreement with [15]. 30-50% selectivity to BTO was detected for
K- and Cs- based samples having M/P≥2 that may be considered as a factor relevant to a faster
deactivation of these catalysts.
The ratio of rates for the elimination-rearrangement routes clearly depends on the nature of
alkali cation and the M/P ratio (Fig. 17B). While at M/P=0.5-1 all fresh samples exhibit comparable
and low E/R values, at higher metal loading E/R ratio increases in the order Na<K≈Cs. All the
Page 25 of 39

catalysts exhibit the highest values of E/R within the same range of M/P~1.5-2. Lower maximum of
E/R~4 is observed for Na P series and higher maximum of E/R~5-9 is found for K P and Cs
_
_
_
P
series (Table 1, Fig. 17B). Therefore, a bulkier metal cation provides better elimination rates. Good
correlation of acid-base pair density (Fig. 9) and E/R ratio (Fig. 17B) for all supported alkali
phosphates may be observed. This correlation favors the idea of E1cB -type elimination mechanism,
where concerted action of acid-base pairs is required.
Comparison of apparent activation energies (Fig. 14, Table 2) shows that an increase of cation
size leads to the decrease of the experimental activation energy both for elimination and
rearrangement pathways. Then not only density of acid-base pairs but their strength is important
factor in the catalytic action. In the case of elimination that requires acid-base pair this effect is
expected. A decrease of the activation energy for rearrangement with increase of cation is not
expected, because Brønsted site is likely to be engaged. Probable explanation of this effect is related
to the flexibility of a cation position around P-O bond. In this case bulkier cation has a better chance
to interact with both oxygens around P atom (Fig. 15B) providing better chances for the proton to
leave i.e., to protonate BDO. Another important feature is low apparent activation energies for
dehydrative epoxidation while the nature of cation seems to be not important.
Reported values of the activation energy for elimination were estimated for thoria (105
kJ/mol) [47] and for amorphous calcium phosphate(233 kJ/mol) [47]. For the rearrangement
pathway in supercritical water apparent activation energy of 147 kJ/mol was reported [48]. DFT
calculation for dehydrative epoxidation of BDO on Cs/SiO gave about 42 kJ/mol activation energy
2
[15]. In general, there is a good correspondence of available data with the data presented here
except the case of calcium phosphate catalyst.
Earlier in the text we pointed out that 10% wt. catalysts have higher density of very weak and
weak acid-base pairs (Fig. 6). Since catalytic data show that both 10% wt. and 20% wt. series of
catalysts have nearly same selectivity patterns, then very weak and weak acid-base sites are not
relevant for the catalytic dehydration of BDO.
Page 26 of 39

During the catalytic run accumulation of dense materials at the exit lines of reactor may occur.
For the basic catalysts with M/P≥2.5, especially for K- and Cs-based catalysts, this effect is much
more painful due to the clogging and raise of back pressure. Since all the catalysts with M∕P≥2.5
have high selectivity to epoxides (Fig. 17A), then this effect of line blocking seems to be related to
the polymerization of BTO. Taking into account the chemistry of epoxide polymerization [49],
strong base catalyzed oxyanionic oligomerization of BTO may be suspected. The polymerization is
followed by secondary reactions of polymer decomposition assuming high temperature inside the
reactor. In the case of homolytic cleavage secondary products of polymer decomposition will have
radical nature. If some of these products are light enough and escape the reactor due to the high rate
of a carrier gas flow, then a radical polymerization may be induced owing to the high concentration
of -olefinic products in the flow (e.g., 3B2OL) on the colder parts of the catalytic system causing
accumulation of dense materials at the exit of reactor.
On the other hand, allyl alcohol may be an initiator in the epoxide base-catalyzed
polymerization [50]. Then taking into account high concentration of olefinic alcohol 3B2OL, base
catalyzed polymerization of epoxides may be initiated by a formation of alkoxides from 3B2OL and
basic sites of the catalyst followed by the polymerization of BTO.
The third issue to consider in the deactivation – the ability of the phosphoric acid to
dehydrogenate hydrocarbons [51] on one hand, and the ability of basic catalysts to dehydrogenate
alcohols [52] on another. Indeed, considerable amount of BDOn and Actn for acidic and basic
catalysts (Table 1) is indicative of dehydrogenation of BDO. From this point of view
dehydrogenation of 3B2OL to acrolein cannot be ruled out. If this happens, then inevitable
polymerization of acrolein will lead to deactivation. The latter hypothesis may be relevant for the
explanation of a faster deactivation of catalysts during 3B2OL conversion (Fig. 13) compared to
BDO conversion (Fig. 11).
5. Conclusions
Silica-supported alkali phosphates at low M/P≤1 form polyphosphate polymers that have
Page 27 of 39

conjugated acid-base pairs and grafted phosphoric acid that represents isolated Brønsted centers.
Acid-base conjugated pairs may be formed by the end groups of polyphosphates having both P-OH
-
+
and P-O M entities in the close proximity. At M/P≥1.5 isolated Brønsted acid centers are probably
formed by P(OH) end groups of short polyphosphates, while acid-base pairs are the same P-OH
2
-
+
and P-O M entities formed as a result of cation migration from pyrophosphates and
tripolyphosphates to silanols of silica.
Catalytic properties suggest that isolated Brønsted acid centers are responsible for the
dehydration of BDO by rearrangement, while conjugated acid-base pairs are responsible for the
dehydration by elimination. Basic centers are active in the dehydrative epoxidation. Both basic and
isolated acid centers also promote dehydrogenation of BDO. For the fresh samples at M/P≤1
rearrangement dominates, but isolated acid centers deactivate fast, and after the deactivation
elimination starts to dominate. At M/P≥1.5 elimination prevails due to the small density of isolated
acid centers and higher density of conjugated acid-base pairs. Elimination/Rearrangement profile of
catalysts follows the pattern of densities of conjugated acid-base pairs with a maximum around
M/P=1.75-2. Higher densities of conjugated acid-base pairs that are observed for cesium and
potassium phosphates supported on silica in turn provide higher E/R ratios for these catalysts
compared to sodium phosphate.
3
B2OL dehydration leads to a faster deactivation compared to BDO dehydration and
deactivation rates are increasing in the row Na<K<Cs. Increase of M/P ratio also leads to the faster
deactivation. Possible reactions causing deactivation are dehydrogenation of 3B2OL leading to
acrolein and/or base catalyzed polymerization of BTO.
Acknowledgments
This work was supported by the Industrial Strategic Technology Development Program (No.
1
0050407) funded by the Ministry of Trade, Industry and Energy (MOTIE, Korea).
Page 28 of 39

Table 1. Catalytic performance in BDO conversion for the fixed bed operation
Carbon balance Elimination Rearrange-
Epoxi-
Dehydroge-
nation
XBDO
,
*
Sample name
in/out, mol/mol SBD+3B2OL
,
ment
E/R
dation
%
%
SIBA+MEK
,
SBTO, %
SBDOn+Actn, %
%
0
1
1
1
2
2
.5Na_P/SiO _20
59
65
78
80
82
35
0.97
0.99
0.95
0.93
0.95
0.77
46
55
64
68
69
22
39
36
31
21
20
44
1.2
1.6
2.1
3.3
3.6
0.5
6
3
1
4
2
9
5
2
Na_P/SiO _20
3
2
.5Na_P/SiO _20
2
2
.75Na_P/SiO _20
3
2
Na_P/SiO _20
6
2
.5Na_P/SiO _20
24
2
0
1
1
1
2
3
0
1
1
1
1
2
2
.5K_P/SiO _20
43
46
39
59
64
87
43
88
73
83
46
67
47
0.69
0.80
0.80
0.90
0.94
0.97
0.94
0.89
0.85
0.89
0.95
0.72
0.95
56
57
71
72
78
35
38
48
48
50
50
26
25
31
25
13
12
9
1.8
2.3
5.4
6.2
9.2
1.1
1.5
1.8
4.7
4.2
5.3
2.2
2.3
4
10
5
7
5
2
K_P/SiO _20
2
.5K_P/SiO _20
9
2
.75K_P/SiO _20
7
7
2
K_P/SiO _20
2
10
2
K_P/SiO _20
33
26
26
11
12
10
12
11
3
22
2
.5Cs_P/SiO _20
23
11
6
6.4
1.2
13.8
17.3
28.6
19.7
48.9
2
Cs_P/SiO _20
2
.25Cs_P/SiO _20
2
.5Cs_P/SiO _20
4
2
.75Cs_P/SiO _20
6
2
Cs_P/SiO _20
5
2
.5Cs_P/SiO _20
9
2
The values in the Table are based on averaged data points during 1-20 h (i.e., 20 h total
averaging) period of the TOS and standard conditions for catalytic runs (0.17 g catalyst, 400°C, 0.2
ml/h BDO, 100 ml/min N2)
*
–
E/R=S(BD+3B2OL+CA+2M2P1OL)/S(IBA+MEK), where CA = crotyl alcohol,
2
M2P1OL = isobutenyl alcohol (abbreviations according to reference [6])
Page 29 of 39