Selectivity and Lifetime Effects in Zeolite-Catalysed Baeyer–Villiger Oxidation Investigated in Batch and Continuous Flow

Article abstract of DOI:10.1002/cctc.201600955

In this manuscript, we investigate the kinetic, mechanistic and lifetime aspects of the Baeyer–Villiger oxidation of cyclohexanone with Sn-β as catalyst and H₂O₂ as oxidant, with the aim of: 1) elucidating the overall reaction network, 2) closing the carbon balance, particularly at high levels of conversion, and 3) examining the intensification of this process in the continuous regime. The results presented herein conclusively demonstrate that this reaction is highly selective for the desired product (?-caprolactone) only below conversions of 60 %. Above this level of conversion, unavoidable hydrolysis of ?-caprolactone to 6-hydroxyhexanoic acid is observed, which consumes the desired product and leads to a reduction in catalytic activity through poisoning. By elucidating the reaction network and working under optimised conditions, we show the potential viability of this methodology to operate continuously over a 180 h period, both at high levels of productivity (324 g (cyclohexanone converted) cm^?3 (reactor volume) kg^?1 (catalyst) h^?1) and selectivity (70 % at 60 % conversion). Over 5000 substrate turnovers were observed during this period, an order of magnitude higher than previously noted for this particular catalyst system.

Full text of DOI:10.1002/cctc.201600955

DOI: 10.1002/cctc.201600955
Full Papers
Selectivity and Lifetime Effects in Zeolite-Catalysed
Baeyer–Villiger Oxidation Investigated in Batch and
Continuous Flow
Keiko Yakabi,[a] Kirstie Milne,[a] Antoine Buchard,[b] and Ceri Hammond*[a]
In this manuscript, we investigate the kinetic, mechanistic and
lifetime aspects of the Baeyer–Villiger oxidation of cyclohexa-
none with Sn-b as catalyst and H2O2 as oxidant, with the aim
of: 1) elucidating the overall reaction network, 2) closing the
carbon balance, particularly at high levels of conversion, and
3) examining the intensification of this process in the continu-
ous regime. The results presented herein conclusively demon-
strate that this reaction is highly selective for the desired prod-
uct (e-caprolactone) only below conversions of 60%. Above
this level of conversion, unavoidable hydrolysis of e-caprolac-
tone to 6-hydroxyhexanoic acid is observed, which consumes
the desired product and leads to a reduction in catalytic activi-
ty through poisoning. By elucidating the reaction network and
working under optimised conditions, we show the potential vi-
ability of this methodology to operate continuously over
a 180 h period, both at high levels of productivity (324 g(cyclo-
hexanone converted)cmÀ3 (reactor volume)kgÀ1 (catalyst)hÀ1
)
and selectivity (70% at 60% conversion). Over 5000 substrate
turnovers were observed during this period, an order of mag-
nitude higher than previously noted for this particular catalyst
system.
Introduction
The Baeyer–Villiger oxidation (BVO) reaction has been widely
studied over the past 100 years, since its discovery in 1899.[1] It
has demonstrated itself to be an especially profitable method
to convert ketones, readily available building blocks in organic
chemistry, into more complex and valuable esters and lac-
tones.[2] A key example of this chemistry is the industrially em-
ployed BVO of cyclohexanone (CyO) to yield e-caprolactone,
a key monomer in the bulk chemical industry.[3] Traditionally,
this chemistry has been performed using peracid-based oxi-
dants, such as mCPBA (meta-chloroperbenzoic acid). Unfortu-
nately, such oxidants possess several drawbacks, including the
co-production of byproducts (carboxylic acids), low active
oxygen content, potential explosiveness and a requirement for
undesirable chlorinated solvents.[4] Additionally, they typically
demonstrate poor levels of selectivity when other functional
groups, such as C=C double bonds, are present in the
substrate.
To overcome these problems, much effort has been made
towards replacing such oxidants with greener, less polluting
and potentially less expensive oxidants.[5] A key example is hy-
drogen peroxide (H2O2).[6] In this case, a suitable catalyst is re-
quired to allow H2O2 to interact with the ketone substrate. Ac-
tivation of the oxidant, in addition to the substrate, may there-
fore be required. Amongst several potential catalytic materials,
the Lewis acidic zeolite, Sn-b, has conclusively been shown by
Corma et al.[3a] to be the most active, selective and suitable cat-
alyst for BVO. This material appears to be uniquely able to po-
larise the carbonyl group of the substrate, thus making it more
prone to nucleophilic attack by the oxidant, H2O2, without acti-
vating the oxidant molecule, thus resulting in high levels of
chemoselectivity even when multiple functional groups are
present. The most widely accepted mechanism for this reaction
is described in Scheme 1.
[a] K. Yakabi, K. Milne, Dr. C. Hammond
Cardiff Catalysis Institute
Although much work has subsequently been devoted to-
wards scalable synthesis of Sn-b,[7] and the development of
novel catalytic processes catalysed by the same (or similar) ma-
terial(s), research focusing on the Sn-b-catalysed BVO reaction
itself is lacking, despite the considerable industrial potential.
Consequently, several questions and challenges remain. Great-
est amongst these are: i) the need to identify and optimise all
the kinetic parameters of the reaction, ii) the requirement to
identify the byproducts, and hence close the low carbon bal-
ance often observed, and iii) investigation of the potential scal-
ability of the system, particularly in the continuous regime.
These three challenges constitute the focus of this work.
Cardiff University
Main Building, Park Place
Cardiff, CF10 3AT (UK)
E-mail: hammondc4@cardiff.ac.uk
Homepage: http://blogs.cardiff.ac.uk/hammond
[b] Dr. A. Buchard
Centre for Sustainable Chemical Technologies (CSCT)
Department of Chemistry
University of Bath
Bath, BA2 7AY (UK)
Supporting information and the ORCID identification number(s) for the
author(s) of this article can be found under http://dx.doi.org/10.1002/
cctc.201600955 and, as per EPSRC requirements, also from http://
www.cardiff.ac.uk/cardiff-catalysis-institute.
ChemCatChem 2016, 8, 1 – 10
1
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Scheme 1. Molecular mechanism proposed for the BVO of cyclohexanone catalysed by Sn-b.
Results and Discussion
Impact of kinetic parameters
According to our optimised synthesis protocol,[7a]
a range of Sn-b catalysts were prepared from a com-
mercially available Al-b precursor material (SiO2/
Al2O3 =38, from Zeolyst). Sn loadings of 1, 2, 5 and
10 wt% were achieved by incorporating Sn into the
vacant framework sites by solid-state incorporation
(SSI) with SnII acetate as the tin source.
The synthetic and spectroscopic information of
these materials was recently extensively described in
a detailed manuscript,[7a] and for brevity will not be
shown here. In general, isomorphously substituted
SnIV sites were almost exclusively observed up to
a loading of 5 wt%, although beyond this loading, in-
active, extra-framework SnxOy sites were also ob-
Figure 1. Catalytic activity of 2Sn-b(38)550 for the BVO of cyclohexanone. Left: Selectivity
to caprolactone and total carbon balance, both as a function of cyclohexanone conver-
sion. Right: Conversion and caprolactone yield versus time.
served. Accordingly, our primary catalyst, henceforth
denoted 2Sn-b(38)550, consisted of 2 wt% Sn incor-
porated into the vacant framework sites of zeolite b
(initial SiO2/Al2O3 =38) by SSI at 5508C, as this load-
ing is consistent with a material possessing highly active,
isolated SnIV sites.
the first 30 minutes of reaction, to 50% after 6 h of reaction. It
is also evident that the catalytic rate rapidly decreases above
conversion levels of 60%. According to this kinetic profile, cap-
rolactone yield very quickly reaches 40%, but then stagnates
and eventually decreases slightly at the highest levels of con-
version. This observation, coupled with the corresponding de-
crease in carbon balance at elevated conversion, strongly
points towards the formation of consecutive byproducts
throughout the reaction period. No additional products could
be detected by GC, which is the most routinely applied analyti-
cal technique in this area of research. The identity of these
byproducts is discussed later in the manuscript.
We first explored the activity of this material for the BVO of
CyO under conditions similar to those reported originally by
Corma et al. (Figure 1).[3a] However, reaction temperatures of
1008C (internal temperature) were employed, instead of 908C.
Under the chosen reaction conditions, 2Sn-b(38)550 is a highly
active catalyst, converting >60% of the CyO reactant in 1 h,
and >90% in <6 h. Initially, the system is highly selective
(90%) for caprolactone, demonstrating this to be the primary
reaction product. However, a remarkable decrease in caprolac-
tone selectivity and total carbon balance is observed as the
conversion increases beyond 60%. Indeed, caprolactone selec-
tivity and total carbon balance decrease from >90% during
We subsequently investigated the impact of the Sn loading
of the catalyst (Figure 2). In these experiments, various samples
&
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
2
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Accordingly, opportunities to improve selectivity by
modifying catalyst composition appear unavailable.
As the reaction temperature is decreased from 100
to 408C, a clear decrease in CyO conversion is ob-
served. Kinetic analysis of the time online data re-
veals that the catalytic reaction fits better as
a second-order reaction under the standard condi-
tions (Figure S1 in the Supporting Information), and
an activation energy of 51.1 kJmolÀ1 was determined.
This barrier is comparable to the value reported by
Boronat et al.[8] for analogous materials under
comparable reaction conditions (45 kJmolÀ1).
As observed in Figure 3 (left), regardless of the re-
action temperature, high (>70%) selectivity to capro-
lactone was only observed when CyO conversion
levels were kept <50–60%, after which point selec-
tivity to caprolactone, and total carbon balance,
drops noticeably. The selectivity trend is clearly com-
parable at all reaction temperatures, within experi-
mental error, and selectivity seemingly only depends
on the extent of conversion. This indicates that the
Figure 2. Catalytic activity of various Sn-b samples containing 2–10 wt% Sn for the BVO
of cyclohexanone. Left: Conversion versus time. Right: Selectivity to caprolactone as
a function of cyclohexanone conversion.
of Sn-b, containing either 2, 5 and 10 wt% of tin,
were tested at a fixed amount of Sn relative to cyclo-
hexanone, that is, the catalyst mass was adjusted so
that the same amount of Sn was employed in the re-
actor (1 mol% relative to CyO). We note here that
the theoretical amount of Sn that can be incorporat-
ed into dealuminated zeolite (SiO2/Al2O3 =38) is ap-
proximately 9 wt%, that is, Sn loadings up to and
beyond the theoretical maximum loading were
tested. Utilisation of high Sn loadings can be benefi-
cial to obtain high space-time yield values and en-
hance process profitability. Furthermore, catalysts
containing different Sn contents may exhibit different
selectivity trends.
As can be seen, reactions performed by using the
same substrate/metal molar ratio with 2Sn-b and
5Sn-b presented almost identical catalytic activities
in terms of conversion (and hence turnover frequen-
Figure 3. Catalytic activity of 2Sn-b(38)550 for the BVO of cyclohexanone at various tem-
peratures. Left: Conversion versus caprolactone selectivity during BVO of cyclohexanone
between 40 and 1008C. Right: Arrhenius plot for 2Sn-b(38)550 for the BVO of cyclohexa-
none (y=10.11431À6146.840x).
cy (TOF)). However, a decrease in activity and TOF
was observed as the Sn loading of the catalyst was
increased to 10 wt%. The decrease in the activity of
this sample (initial (5 min) TOF decreasing from
370 hÀ1 (2Sn-b) to 199 hÀ1 (10Sn-b)) indicates that spectator
Sn species are present at higher loadings, in agreement with
our recent studies of transfer hydrogenation and glucose iso-
merisation over the same materials.[7] Accordingly, all kinetic
parameters were optimised in the presence of 2Sn-b so as to
avoid the potential influence of spectator Sn sites.
side reactions or consecutive reactions have comparable acti-
vation energies to the BVO reaction, and that high(er) levels of
selectivity cannot be obtained through kinetic control.
To determine the effect of H2O2 concentration on catalytic
rate, a series of experiments with different H2O2/ketone ratios
were performed, with all other parameters, for example, Sn
content (1 mol%), reaction temperature (1008C) and catalyst
choice (2Sn-b(38)550) remaining unchanged. However, as H2O2
is used as an aqueous solution, the concentration of H2O also
changes during these experiments. The initial concentration of
hydrogen peroxide was varied from 0.035 to 1m, representing
a range of H2O2/ketone of ratios 0.1 to 3.0 (under typical condi-
tions, a ratio of 1.5 was employed). The apparent rate constant
At all levels of conversion, each catalyst exhibits comparable
selectivity, indicating that the formation of spectator Sn sites
does not adversely affect the reaction network, nor catalyse
the formation of byproducts or consecutive reaction products.
Furthermore, the residual Brønsted acid sites from unreacted
silanol nests do not participate in the reaction, indicating that
conversion alone determines overall caprolactone selectivity.
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
3
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
Identification of byproducts and closing the carbon
balance
To fully identify the reaction network, we subse-
quently turned our attention to identifying the
nature of the unknown byproducts that must be
present, accounting for the missing carbon balance.
From previous studies, the most likely byproducts in-
clude 6-hydroxyhexanoic acid (6-HHA), formed
through caprolactone hydrolysis, and adipic acid
(AA), formed by the oxidation of 6-HHA.[9] Although
we were able to record the FTIR spectra of four po-
tential reactant species (CyO, caprolactone, 6-HHA
and AA, Figure S3 in the Supporting Information), the
relative overlap of each C=O signal means that iden-
tification of each potential compound in a mixture of
all compounds was not feasible (Figure S4 in the
Supporting Information).
Figure 4. Left: Effect of [H2O2] on the rate of the BVO of CyO at 1008C. Right: Impact of
relative Sn content on the rate of BVO at 1008C.
Although the presence of the solvent inhibits iden-
tification of all potential species within the reaction
1
solution by H NMR spectroscopy under normal con-
(kapp, units mÀ1 sÀ1) at each H2O2/ketone ratio was determined
at the very initial stages of the reaction, that is, in the linear
region of the kinetic plot, and are represented in Figure 4. Be-
tween H2O2/ketone ratios of 0.1 and 0.75, a first-order depend-
ence on H2O2 is observed, with kapp increasing linearly with
[H2O2]. However, above this H2O2/ketone ratio, kapp reaches
a plateau, within experimental error, and the reaction becomes
zero order with respect to the oxidant. Once in the zero-order
regime, excess H2O2 does not interfere with the intrinsic kinet-
ics of the system. However, at the highest oxidant concentra-
tion, caprolactone selectivity at all levels of ketone conversion
drops somewhat (Figure S2 in the Supporting Information).
This indicates that H2O2 and/or H2O (as H2O2 is used as an
aqueous solution) may be involved in a consecutive reaction
that decreases the total carbon balance.
Under the conditions reported by Corma et al.,[3a] Sn con-
tents of 1 mol% (relative to CyO) are typically employed. Ac-
cordingly, the mass of 2Sn-b(38)550 was varied so that Sn con-
tents of 0.1–3 mol%, relative to CyO, were utilised to deter-
mine the impact of Sn content. The kapp (mÀ1 sÀ1) value at all Sn
contents was calculated at low conversions to minimise the
impact of product-induced deactivation. Between 0.1–1 mol%
Sn, a linear relationship between kapp and Sn content is ob-
served. However, at very high Sn loadings, kapp is lower than
anticipated, indicating that some mass transfer limitations are
present when large masses of catalyst are used. Accordingly,
all further experiments were conducted at Sn contents of
1 mol% or lower, consistent with the reaction being in the ki-
netic regime (also evident from the linear Arrhenius plot). In
terms of caprolactone selectivity, the trend obtained was main-
tained throughout all experiments, that is, varying the Sn con-
tent within the reactor did not overtly impact caprolactone se-
lectivity at any level of conversion.
ditions, the relatively high boiling points of CyO and all the po-
tential products allowed us to almost completely evaporate
the solvent (1,4-dioxane, boiling point 1008C) prior to analysis,
and consequently obtain satisfactory qualitative data. This im-
mediately revealed the presence of distinct resonances of CyO,
caprolactone and 6-HHA in various reaction samples. Unfortu-
1
nately, all H NMR resonances of adipic acid overlap with CyO
and 6-HHA (Figures S5 to S8 in the Supporting Information),
and hence AA cannot be conclusively identified by 1H NMR
spectroscopy.
1
Quantification of the (by)products by H NMR spectroscopy
was achieved by calibrating a capillary containing a known
amount of tetramethylsilane (s, d=0.0 ppm) against known
standards (Figure S9 in the Supporting Information), and plac-
ing this capillary within the NMR tubes of various time online
samples. To verify this approach, the concentrations of capro-
lactone as determined by GC-FID (biphenyl internal standard)
and 1H NMR spectroscopy were compared. All values were
found to be identical within 1–6% error, verifying this ap-
proach (Table S1 in the Supporting Information). We note here
that overlapping signals prevent quantification of CyO by this
method, and its concentration was solely monitored by GC-
1
FID. Utilising both GC-FID and H NMR spectroscopy, the initial
BVO of CyO was repeated under standard conditions. With the
added quantification of 6-HHA, the carbon balance improves
from 54% to 88% at the highest levels of conversion
(Figure 5). This clearly demonstrates that as the conversion in-
creases, selectivity towards 6-HHA increases dramatically,
strongly indicating that the overall reaction contains two par-
ticular steps: 1) BVO, yielding caprolactone, and 2) hydrolysis
of caprolactone to 6-HHA.
To determine the concentration of AA, which may be pres-
ent as a deep oxidation product, 13C NMR spectroscopy and
HPLC were utilised; authentic standards revealed that AA
could be separated and quantified by HPLC, and that it pos-
sesses a distinct 13C resonance at 175.82 ppm. Although Cavani
&
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
4
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
Figure 5. Time online analysis of BVO of CyO at 1008C monitored by GC and
1H NMR, catalysed by 2Sn-b(38)550
.
1
Figure 6. Time online analysis of BVO of CyO at 1008C by H NMR spectros-
copy.
et al.[10] observed AA as a deep oxidation product during the
BVO of CyO with titanium silicalite-1 as catalyst, only trace
amounts (<3 mol%) of this byproduct were observed during
the entire reaction period (Figure S10 in the Supporting Infor-
mation). We attribute the minor amount of this deep oxidation
product in this particular to 1) the change in catalyst utilised
gomer formation. The degree of caprolactone oligomerisation
likely fully closes the carbon balance at high levels of conver-
sion, although precise quantification of the oligomeric
products is not possible at this stage.
C
(titanium silicalite-1 is a catalyst known to produce OH radi-
cals, which may readily oxidise 6-HHA), and 2) the presence of
C
Caprolactone conversion studies
1,4-dioxane as solvent, which acts as an OH radical
scavenger.[10]
In an attempt to improve the overall caprolactone selectivity
levels—particularly at high levels of conversion—the caprolac-
tone hydrolysis reaction, responsible for the co-production of
6-HHA, was also studied. Although dehydration of 6-HHA to
caprolactone could be performed in a second step, obtaining
high caprolactone selectivity would lead to improved efficiency
and minimise downstream processing.
To further close the carbon balance (91% including AA), the
spent catalytic materials extracted from the reactor after reac-
tion were also analysed by thermogravimetric analysis (TGA;
Figure S11 in the Supporting Information, including detailed
experimental description). The mass losses observed upon
heating the ex reactor catalyst were compared with those mass
losses observed from known standards of CyO, caprolactone,
6-HHA and AA, pre-adsorbed onto a reference 2Sn-b(38)550
powder. Thus, TGA of the used catalyst allowed us to deter-
mine the type (from the desorption temperature) and quantity
(by the mass loss observed at said temperature) of adsorbed
carbon present. In addition to some residual solvent, a consid-
erable amount (>11 wt%) of caprolactone, and a small
amount (3 wt%) of 6-HHA were detected. In total, this corre-
sponded to approximately 5% of the original carbon species
introduced into the reaction at the start of the process, and
thus raises the carbon balance under the typical BVO
conditions to 96%.
We first investigated the role of the catalyst in the hydrolysis
reaction. To do this, a solution of caprolactone in 1,4-dioxane
(0.2m) was placed in a round-bottomed flask under the stan-
dard reaction conditions, and two separate experiments were
performed. In the first reaction (Table 1, entry 1), the typical
amount of H2O (added by necessity when using 50 wt% H2O2)
Table 1. Extent of caprolactone hydrolysis observed under various
reaction conditions.
Entry
Catalyst
H2O
[mL]
Conversion
[%]
Over the first hour of the reaction, no oligomeric products
of caprolactone and/or 6-HHA, were detected by NMR spec-
troscopy or gel permeation chromatography (GPC). However,
at more extended reaction times (2–6 h), and higher levels of
conversion, some polymeric products could be observed
(Figure 6). GPC analysis of these reaction samples revealed
these polymeric species to be oligomers of caprolactone, pos-
sessing between two and three caprolactone units (Figure S12
in the Supporting Information). This clearly suggests that per-
forming the reaction continuously at high conversion should
be avoided so as to minimise the opportunity of fouling by oli-
1
2
3
4
5
6
7
8
–
0.15
0.15
0.15
0.15
0.15
0.3
0
0
0
25
1.8
23
13
36
1.1
8
7
4
4
62
2% Sn-b 38
deAl b 38
5% Sn-b 38
10% Sn-b 38
2%Sn-b 38
deAl b 38
2% Sn-b 38
5% Sn-b 38
10% Sn-b 38
2%Sn-b 38 recalcined
2% Sn-b 38
9
0
0
0
10
11
12
0.3 (50 wt% H2O2)
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
5
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
was added to the vessel. In the second experiment (entry 2),
2Sn-b was also added to the reactor, in addition to H2O. This
control experiment was to verify whether hydrolysis of capro-
lactone proceeded in an uncatalysed manner, or whether the
heterogeneous material promoted the reaction. In both cases,
consumption of caprolactone and the formation of 6-HHA was
monitored by GC and 1H NMR spectroscopy (Table S2 in the
Supporting Information). In the absence of catalyst (Table 1,
entry 1), no conversion of caprolactone was observed. In con-
trast, in the presence of the catalyst, 25% conversion of capro-
lactone to 6-HHA (at >95% selectivity) was observed. This
strongly indicates that some catalyst functionality (potential
Lewis acidity, Brønsted acidity, or defect sites) is required to
catalyse the hydrolysis reaction.
H2O2 solution, both of which may increase the rate of ring
opening. Finally, we investigated the temperature dependence
of the hydrolysis reaction catalysed by Sn-b, obtaining an Eact
of 37 kJmolÀ1 (Figure S18 in the Supporting Information), com-
parable to that of the BVO reaction. As the reactions are con-
secutive, effort should be made to control the extent of con-
version, and to use the conversion versus selectivity relation-
ship to maximise the selectivity to caprolactone.
Intensification and catalyst stability in continuous flow
Having identified the nature of the byproducts present during
BVO, we investigated the potential of each of the (by)products
to poison the catalyst (product-induced poisoning), as this
may prohibit continuous operation. Indeed, the large decrease
in catalytic rate observed after 1 h of batch reaction (Figure 1)
indicates product-induced poisoning being present. According-
ly, we repeated the standard BVO reaction at 1008C (Figure 1)
in the presence of various quantities of the reaction (by)prod-
ucts (particularly caprolactone, water and 6-HHA; AA was not
studied owing to its low concentration throughout the reac-
tion period). In each case, the conversion of CyO was used to
determine the kinetics of the reaction as a function of the
amount of (by)product added. We note here that as H2O2 is
used as an aqueous solution (50 wt%), a 2.83:1 mole ratio (i.e.,
283 mol%) of H2O relative to CyO is unavoidably present at
the beginning of the reaction. However, assuming a totally se-
lective oxidation process occurs, an additional 150 mol% can
form during the reaction. In these studies, therefore, the
amount of additional water added to the vessel represents this
additional 150 mol%, that is, at 0 mol% H2O byproduct added,
the initial 283 mol% water is still present.
To deduce the nature of the catalyst functionality responsi-
ble for the hydrolysis reaction, the activity of various samples
of Sn-b for the hydrolysis reaction alone was explored. In the
absence of Sn (entry 3), dealuminated zeolite b demonstrates
extremely low reactivity for the hydrolysis reaction, indicating
that defect sites and residual Brønsted acid sites of the zeolite
framework alone are not responsible for caprolactone hydroly-
sis. In contrast, Sn-b (possessing between 2 and 10 wt% Sn,
entries 2, 4 and 5) demonstrates substantial activity for hydrol-
ysis, with up to 25% conversion being observed over a four-
hour reaction period. Interestingly, the relative activities of the
2, 5 and 10Sn-b samples for caprolactone hydrolysis is identi-
cal to the relative activities observed with these samples for
the BVO reaction itself (Figure 2). Unfortunately, this strongly
indicates that the same Sn sites required for BVO are responsi-
ble for caprolactone hydrolysis. Thus, catalyst modification
methodologies are unlikely to be able to remove the hydroly-
sis side reaction, and 6-HHA formation is likely unavoidable
under these reaction conditions. Quantification by using the
tetramethylsilane capillary confirmed 6-HHA as the main prod-
uct with a selectivity >90% in all of these cases, although
trace amounts of oligomers could also be observed at extend-
ed reaction periods (Figure S13 in the Supporting Information).
Surprisingly, even in the absence of additional water, conver-
sion of up to 8% caprolactone to 6-HHA was still observed
when using different Sn-b catalysts (entries 7 to 10). We subse-
quently repeated this series of experiments utilising freshly cal-
cined Sn-b as opposed to Sn-b that had been stored at ambi-
ent conditions for approximately 4–7 days (entry 11). Although
caprolactone conversion was reduced, it was not eliminated
completely, showing that even trace amounts of water lead to
hydrolysis.
As seen in Figure 7, each of the products influences the ini-
tial rate of reaction to a particular extent. Caprolactone, the de-
sired product, leads to a decrease in the initial rate, but only
by approximately 20% even at relatively high, that is,
50 mol%, concentrations. Water and 6-HHA, on the other
hand, strongly influence the initial catalytic activity. The initial
Additionally, we studied the effect of H2O2 addition
(entry 12). Whilst higher caprolactone conversion was observed
when H2O2 is present alongside H2O (62% (entry 12) versus
36% (entry 6)), AA was still not observed by 13C NMR spectros-
copy (Figure S14 in the Supporting Information), and selectivity
to 6-HHA was still >90%. The high selectivity to 6-HHA ob-
served even in this case strongly indicates that ring opening,
and not deep oxidation, remains the dominant consecutive re-
action. We hypothesise that the increased rate of ring opening
in the presence of H2O2 may be due to the lower pH/pKa of
the solution and/or the presence of cationic stabilisers in the
Figure 7. Effect of byproduct addition on the initial rate of BVO observed
with 2Sn-b(38)550. Inverse rate constants were determined by measuring the
decrease in CyO concentration as a function of reaction time by GC-FID.
&
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
6
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
rate of reaction is found to decrease by approximate-
ly 50% when 50 mol% 6-HHA, relative to CyO, is
added, and water inhibits the rate by approximately
40% at a byproduct level of 50 mol% (which corre-
sponds to a H2O2 conversion level of approximately
33%, assuming total conversion of H2O2 to H2O
during the reaction). This clearly demonstrates that
the continuous reactor should be held at a level of
conversion consistent with there being as little 6-
HHA and H2O present as possible. A CyO conversion
level of approximately 50–60% (corresponding to
a 6-HHA level of approximately 15–25 mol%, and
a decrease in rate of 20%) appears to be a good
target. The strong decrease in rate observed at 6-
HHA levels of 50 mol% corresponds well to the de-
crease in catalytic rate observed in the initial batch
experiments at conversion levels of 60% and above,
where 6-HHA begins to become the major reaction
product.
Figure 8. Left: Conversion versus time on stream during BVO of CyO in the continuous
regime at 1008C. Right: Selectivity to caprolactone observed at different degrees of CyO
conversion in both batch and continuous mode.
Having deduced the impact of the potential reac-
tion products on reaction rate, we subsequently set
about exploring the activity of this system in continuous flow.
A continuous plug flow reactor (PFR) offers major advantages
over a slurry reactor, including i) improved process control and
safety, ii) excellent mass and heat transfer, iii) shorter reaction
times, iv) smaller reactor volumes, iv) scalability and v) in-
creased space-time yields. In addition, continuous PFRs allow
critical evaluation of a catalyst’s stability and lifetime under
steady-state conditions. The reactor employed was adapted
from a kinetically relevant PFR optimised for the Sn-b catalysed
isomerisation of glucose to fructose[11] and the transfer hydro-
genation of carbonyl compounds.[12] The kinetic relevance of
the reactor for BVO was also fully benchmarked (Figure S19 in
the Supporting Information), with testing being performed in
the absence of external mass transfer limitations.
initial activity (not shown), demonstrating that the slight de-
crease in activity is reversible, that is, deactivation is not
permanent.
In addition to high levels of stability, performing the reaction
in plug flow mode also leads to tremendous improvements in
the productivity of the system, even at the same reaction tem-
peratures used for the batch experiments. The space-time
yield (STY) at 1008C in the continuous reactor increases by one
order of magnitude relative to the batch reactor system
(Table 2).[14] Notably, the caprolactone selectivity obtained at
the maximum STY of each reactor is also substantially higher in
the PFR (70 vs. 39%), although as a function of conversion no
substantial improvements are observed at the same H2O2/
ketone ratio.
According to the product-induced deactivation studies
above (Figure 7), we targeted contact times that would lead to
a conversion level of approximately 60% so as to minimise the
influence of 6-HHA on catalyst lifetime. Initially, a H2O2/ketone
ratio of 1.5 was employed, consistent with typical batch experi-
ments. Apart from a very short induction period (Figure 8, left,
blue triangles), the catalytic activity remains relatively constant
over a period of 180 h, with a steady-state CyO conversion of
60% being observed. No major modifications in selectivity
were observed upon moving into the continuous reactor, and
the conversion versus caprolactone selectivity trend sits on the
same profile as found for the batch reactor at the same H2O2/
ketone ratio (Figure 8, right).
Table 2. Relative performance of Sn-b for BVO chemistry in batch and
flow.
Reactor
type
Space-time
yield[a]
Relative space-
time yield
Lactone selectivity @ max.
STY [%]
batch[b]
flow[c]
24.5
323.9
1
13.2
39
70
[a] Space-time yields were calculated at maximum conversion as grams of
reactant converted per cm3 reactor volume, per hour and per kg of cata-
lyst. [b] Only the liquid volume was used as the reactor volume.
[c] Volume of catalyst bed (including diluent) used as the reactor volume.
Clearly, the Sn-b/H2O2 BVO system possesses good levels of
stability. Over the course of 180 h, the catalyst performed
>5000 substrate turnovers,[13] equivalent to over 50 complete
batch reactions. Notably, these TON values are an order of
magnitude higher than previously observed for the same cata-
lytic system.[3a] Overall, a relative loss of only 15.4% of maximal
activity (from 65 to 55%) was observed during this period,
even without periodic regeneration being performed. Recalci-
nation of the catalyst bed after 180 h nevertheless restores the
We subsequently explored the potential of the system to
operate in an oxidant-limited regime. From the batch studies
(Figure 4, left), it was shown that similar catalytic rates were
observed at a H2O2/ketone ratio of 0.75 as were observed at
the typical 1.5 ratio, that is, the BVO reaction is zero order in
[H2O2] above this level. Although operating in an oxidant-limit-
ed regime would limit the system to <75% CyO conversion,
the target level of conversion (50–60%) would still be achieva-
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
7
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

Full Papers
ble, and the decrease in water may limit the caprolactone hy-
drolysis reaction and product-induced poisoning. We subse-
quently repeated the continuous experiment at the lower
H2O2/ketone ratio of 0.75, keeping all other experimental pa-
rameters, including temperature, flow rate, catalyst mass and
reactor volume, identical. A high level of stability is observed
in the oxidant-limited regime, although CyO conversion was
approximately 10% lower across the entire time period. In
terms of activity, the maximum conversion obtained was 51%,
dropping to 44% after 180 h, which represents a relative de-
crease from maximal activity of 13.7%. As such, stability is very
slightly improved in the oxidant-limited regime. This may be
due to the lower amount of water, or the slightly lower level
of conversion in this system, which leads to less 6-HHA forma-
tion. Similar improvements in caprolactone selectivity at all
overlapping ketone conversion levels were also observed in
the oxidant-limited regime, with caprolactone selectivity in-
creasing to 85% at a CyO conversion of 40%. However, given
the minimum amount of H2O required for ring opening, it is
clear that complete suppression of ring opening under these
conditions is unlikely.
converted)cmÀ3 (reactor volume)gÀ1 (catalyst)hÀ1. In addition
to being one order of magnitude more productive on a space-
time-yield basis, the continuous system is also more selective
for caprolactone at these elevated productivities (70% vs.
39%), and more selective with the H2O2 utilised, thus demon-
strating the major advantages of the continuous system.
Experimental Section
Catalyst synthesis and pre-treatment
Zeolite Al-b (Zeolyst, NH4-form, SiO2/Al2O3 =38) was dealuminated
by treatment in HNO3 solution (13m HNO3, 1008C, 20 h, 20 mLgÀ1
zeolite). The dealuminated powder was washed with water and
dried overnight at 1108C. Solid-state stannation was performed by
grinding the appropriate amount of tin(II) acetate with the neces-
sary amount of dealuminated zeolite for 10 minutes in a pestle
and mortar, prior to heat treatment in a combustion furnace (Car-
bolite MTF12/38/400). A two-stage heat treatment was employed:
to 5508C (108CminÀ1 ramp rate) first in a flow of N2 (3 h not in-
cluding ramp period) and subsequently in air (3 h) for a total of 6 h
plus heating and cooling periods. Gas flow rates of 60 mLminÀ1
were employed at all times. The sample was held horizontally in an
alumina combustion boat (10 mL capacity), and a quartz tube was
used to seal the sample environment.
Operating with less oxidant also leads to major improve-
ments in the H2O2-based selectivity of the reaction (Figure S20
in the Supporting Information). At a H2O2/ketone ratio of 0.75,
a H2O2-based selectivity (i.e., moles CyO converted/moles H2O2
consumed) of 85Æ4% was obtained, clearly indicating that
most of the peroxide is used selectivity for oxidation. In con-
trast, a maximum H2O2-based selectivity of only 57% was ob-
tained when a H2O2/ketone ratio of 1.5 was employed under
otherwise identical conditions. Clearly, decreasing the concen-
tration of H2O2 leads to noticeable improvements in H2O2 uti-
lisation, without overly compromising the BVO process itself.
Such improvements are of tremendous importance to both
process economics and the sustainability of the reaction,
particularly given the cost of H2O2.
Kinetic evaluation and analytical methods
Batch BVO reactions were performed in a 100 mL round-bottomed
flask equipped with a water-free reflux condenser. The reaction
temperature was controlled by immersion in a silicon oil bath. The
vessel was charged with a solution of cyclohexanone in 1,4-diox-
ane (10 mL, 0.33m), which also contained an internal standard (bi-
phenyl, 0.01m), and the appropriate amount of catalyst. The vessel
was subsequently heated to the desired temperature (1008C inter-
nal temperature). The reaction was initiated by addition of an ap-
propriate amount of H2O2, typically corresponding to a H2O2/
ketone ratio of 1.5. The solution was stirred at 750 rpm with an
oval magnetic stirrer bar. Aliquots of the reaction solution were
taken periodically for analysis, and were centrifuged prior to injec-
tion into a GC (Agilent 7820, 25 m CP-Wax 52 CB). Reactants were
quantified against a biphenyl internal standard. H2O2 concentra-
tions were determined by titration against Ce4+. Batch caprolac-
tone hydrolysis reactions were performed by using the same pro-
cedure and reaction conditions as used for the BVO reaction,
although cyclohexanone was replaced with caprolactone.
Conclusions
The Baeyer–Villiger oxidation of cyclohexanone by Sn-b/H2O2
was studied in batch and continuous mode, in an attempt to
understand the mechanistic aspects of the reaction, and identi-
fy its potential scalability. It is demonstrated that the BVO of
CyO by Sn-b/H2O2 is a challenging reaction, and that consecu-
tive reactions—particularly the hydrolysis of the primary capro-
lactone product—can substantially decrease the efficiency (se-
lectivity) and activity (rate) of the process. Largely, this is due
to the unavoidable formation of 6-hydroxyhexanoic acid (6-
HHA) by caprolactone hydrolysis, which leads to partial poison-
ing of the Sn-b catalyst, and also consumes the desired capro-
lactone product, decreasing the carbon-based selectivity. Opti-
mising the reaction conditions so as to minimise 6-HHA forma-
tion nevertheless shows the potential scalability of this system
in the continuous regime. Operating the reaction at conversion
levels <60%, at which point 6-HHA formation is less pro-
nounced, results in a stable catalytic system that operates for
180 h without major loss in activity, yielding a turnover
number >5000, and a volumetric productivity of 324 g(ketone
Continuous BVO reactions were performed in a home-made plug
flow, stainless steel, tubular reactor. Reactant delivery (0.33m CyO
in 1,4-dioxane, H2O2/ketone 1.5 or 0.75) was performed by an HPLC
pump. The catalyst (0.4 g) was mixed with a diluent material (SiC
(particle size of 63–75 mm), 1.6 g) to avoid back mixing and to min-
imise the pressure drop, and the bed was placed in between two
plugs of quartz wool. The diluted catalyst was densely packed into
1
a = inch stainless steel tube (4.1 mm ID), and a frit (0.5 mm) was
4
placed at the end of the bed to avoid any loss of material. A con-
tact time of 9.75 min was typically employed. The reactor tempera-
ture was controlled by immersion in a thermostatted oil bath, and
the pressure was controlled by means of a backpressure regulator.
The reaction feed was identical to that used for batch reactions.
Aliquots of the BVO reaction solutions were taken periodically
from a sampling valve placed after the reactor, and were analysed
in the same manner as the batch reactions.
&
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
8
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!

Full Papers
93–100; d) Z. Q. Lei, Q. H. Zhang, J. J. Luo, X. Y. He, Tetrahedron Lett.
2005, 46, 3505–3508; e) J. Dijkmans, W. Schutyser, M. Dusselier, B. F.
Sels, Chem. Commun. 2016, 52, 6712–6715.
Acknowledgements
C.H. designed and guided the research. K.Y. prepared the cata-
lysts, and K.Y. and K.M. tested the catalysts for under the supervi-
sion of C.H. Analytical experiments were devised and guided by
C.H. and A.B., and were performed by K.Y. All authors have given
approval to the final version of the manuscript. C.H. is grateful to
The Royal Society for research grant funding (RG140754) and
provision of a University Research Fellowship (UF140207). K.Y.
and K.M. thank the EPSRC for Ph.D. studentship funding through
the Catalysis Centre for Doctoral Training. A.B. thanks Roger and
Sue Whorrod for a Whorrod Research Fellowship at the CSCT.
[7] a) C. Hammond, D. Padovan, A. Al-Nayili, P. P. Wells, E. K. Gibson, N. Di-
mitratos, ChemCatChem 2015, 7, 3322–3331; b) P. Wolf, C. Hammond,
S. Conrad, I. Hermans, Dalton Trans. 2014, 43, 4514–4518; c) W. N. P. van
der Graaff, G. Li, B. Mezari, E. A. Pidko, E. J. M. Hensen, ChemCatChem
2015, 7, 1152–1160; d) J. Dijkmans, D. Gabriels, M. Dusselier, F. de Clip-
pel, P. Vanelderen, K. Houthoofd, A. Malfliet, Y. Pontikes, B. F. Sels, Green
Chem. 2013, 15, 2777–2785; e) P. Y. Dapsens, C. Mondelli, J. Jagielski, R.
Hauert, J. P. Ramirez, Catal. Sci. Technol. 2014, 4, 2302–2311.
[8] M. Boronat, A. Corma, M. Renz, J. Sastre, P. M. Viruela, Chem. Eur. J.
2005, 11, 6905–6915.
[9] a) R. A. Steffen, S. Teixeira, J. Sepulveda, R. Rinaldi, U. Schuchardt, J. Mol.
Catal. A 2008, 287, 41–44; b) P. Li, G. Liu, H. Wu, Y. Liu, J. Jiang, P. Wu, J.
Phys. Chem. C 2011, 115, 3663–3670; c) C. Xia, L. Ju, Y. Zhao, H. Xu, B.
Zhu, F. Gao, M. Lin, Z. Dai, X. Zou, X. Shu, Chin. J. Catal. 2015, 36, 845–
854.
Keywords: Baeyer–Villiger oxidation · continuous flow · Lewis
acids · tin · zeolites
[10] F. Cavani, K. Raabova, F. Bigi, C. Quarantelli, Chem. Eur. J. 2010, 16,
12962–12969.
[11] D. Padovan, C. Parsons, M. Simplicio, C. Hammond, Green Chem. 2016,
18, 5041–5049.
[12] A. Al-Nayili, K. Yakabi, C. Hammond, J. Mater. Chem. A 2016, 4, 1373–
1382.
[13] Flow rate=0.16 mLminÀ1
; [CyO]=0.33m; mass (Sn-b)=400 mg; Sn
loading=2 wt%. Accordingly, 0.783 moles CyO/mole Sn passes through
the column per minute. Over 180 h, this equals approximately 8456.4
turnovers in the column. At a steady-state conversion of 60% (approxi-
mate average over 180 h) this equals approximately 5074 turnovers.
[14] a) Batch reactor. [CyO]=0.33m; volume=10 mL; mass of Sn-b=
200 mg; max. conversion=91%; time=6 h. b) Flow reactor. [CyO]=
0.33m; flow rate=0.16 mLminÀ1; reactor volume: 1.56 mL; mass of Sn-
b=400 mg; max. conversion=65%.
[1] A. Baeyer, V. Villiger, Ber. Dtsch. Chem. Ges. 1899, 32, 3625–3633.
[2] a) M. Renz, B. Meunier, Eur. J. Org. Chem 1999, 1999, 737–750; b) C. Ji-
mꢁnez-Sanchidriꢂn, J. R. Ruiz, Tetrahedron 2008, 64, 2011–2026.
[3] a) A. Corma, L. T. Nemeth, M. Renz, S. Valencia, Nature 2001, 412, 423–
425; b) C. Hammond, S. Conrad, I. Hermans, Angew. Chem. Int. Ed. 2012,
51, 11736–11739; Angew. Chem. 2012, 124, 11906–11909.
[4] a) G. Strukul, Angew. Chem. Int. Ed. 1998, 37, 1198–1209; Angew. Chem.
1998, 110, 1256–1267; b) M. Renz, T. Blasco, A. Corma, V. Fornes, R.
Jensen, L. Nemeth, Chem. Eur. J. 2002, 8, 4708–4717.
[5] a) T. Katsuki, Chem. Soc. Rev. 2004, 33, 437–444; b) G. J. ten Brink,
I. W. C. E. Arends, R. A. Sheldon, Chem. Rev. 2004, 104, 4105–4124; c) Y.
Imada, H. Iida, S. Murahashi, T. Naota, Angew. Chem. Int. Ed. 2005, 44,
1704–1706; Angew. Chem. 2005, 117, 1732–1734.
[6] a) A. Corma, M. T. Navarro, L. Nemeth, M. Renz, Chem. Commun. 2001,
2190–2191; b) A. Corma, M. T. Navarro, M. Renz, J. Catal. 2003, 219,
242–246; c) U. R. Pillai, E. J. Sahle-Demessie, J. Mol. Catal. A 2003, 191,
Received: August 3, 2016
Published online on && &&, 0000
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
9
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
&
These are not the final page numbers! ÞÞ

FULL PAPERS
K. Yakabi, K. Milne, A. Buchard,
C. Hammond*
&& – &&
Selectivity and Lifetime Effects in
Zeolite-Catalysed Baeyer–Villiger
Oxidation Investigated in Batch and
Continuous Flow
Selectivity and lifetime effects: The
mechanistic and intensification aspects
of the Baeyer–Villiger oxidation of cyclo-
hexanone to caprolactone are investi-
gated under batch and continuous-flow
regimes.
&
ChemCatChem 2016, 8, 1 – 10
www.chemcatchem.org
10
ꢀ 2016 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
ÝÝ These are not the final page numbers!