Organic synthesis by Twin Screw Extrusion (TSE): Continuous, scalable and solvent-free

Article abstract of DOI:10.1039/c6gc03413f

Mechanochemistry provides a method to reduce or eliminate the use of solvents by carrying out reactions through the grinding of neat reagents. Until recently a significant drawback of this form of synthesis has been the limited ability to scale up. Howeve

Full text of DOI:10.1039/c6gc03413f

Green Chemistry
View Article Online
View Journal
PAPER
Organic synthesis by Twin Screw Extrusion (TSE):
continuous, scalable and solvent-free†
Cite this: DOI: 10.1039/c6gc03413f
Deborah E. Crawford,*a Clodagh K. G. Miskimmin,a Ahmad B. Albadarin,b
Gavin Walkerb and Stuart L. James*a
Mechanochemistry provides a method to reduce or eliminate the use of solvents by carrying out reactions
through the grinding of neat reagents. Until recently a significant drawback of this form of synthesis has
been the limited ability to scale up. However, it has been shown that twin screw extrusion (TSE) may over-
come this problem as demonstrated in the continuous synthesis of co-crystals, Metal Organic
Frameworks (MOFs) and Deep Eutectic Solvents (DES), in multi kg h−1 quantities. TSE has provided a
means to carry out mechanochemical synthesis in a continuous, large scale and efficient fashion, which is
adaptable to a manufacturing process. Herein, we highlight the potential of this technique for organic
synthesis by reporting four condensation reactions, the Knoevenagel condensation, imine formation,
aldol reaction and the Michael addition, to produce analytically pure products, most of which did not
require any post synthetic purification or isolation. Each reaction was carried out in the absence of sol-
vents and the water byproduct was conveniently removed as water vapour during the extrusion process
due to the elevated temperatures used. Furthermore, the Knoevenagel condensation has been studied in
detail to gain insight into the mechanism by which these mechanochemical reactions proceed. The
results point to effective wetting of one reactant by another as being critical for these reactions to occur
under these reaction conditions.
Received 12th December 2016,
Accepted 3rd January 2017
DOI: 10.1039/c6gc03413f
www.rsc.org/greenchem
Temperature normally changes incidentally due to the friction
during the milling process.13 Takacs reports that planetary ball
Introduction
Mechanochemical synthesis, conducted by grinding two or milling can result in temperatures exceeding 200 °C, and there
more solid reagents together to instigate a chemical reaction, is currently very limited control of this temperature rise which
is of increasing interest for reducing or eliminating the use of can have a dramatic influence on the overall process.13
hazardous and toxic substances in chemical synthesis and Cryomills can be used, which allow milling to be conducted at
materials processing.1–5 Avoiding toxic solvents not only −196 °C. However this approach is not suitable for all milling
decreases potential hazards, but can also reduce the overall investigations. Alternatively, the ball mill jars can be preheated
cost of a process.5 A variety of materials and compounds have or modified to allow for thermal bath fluids to be circulated
been prepared mechanochemically including metal alloys,6 around a jacket to regulate both high and low temperatures.13
metal organic frameworks (MOFs),7 and organic compounds Overall, however, good temperature control in ball milling is
from reactions such as the Knoevenagel condensation8–10 and currently not widely established. With regard to scalability,
Wittig reaction,11 amongst others.12
typically, the maximum quantity of the product obtained is of
However, two main drawbacks of mechanochemical syn- the order of several hundred grams, through the use of plane-
thesis have been the general lack of both temperature control tary ball mills.9 Larger industrial scale ball mills are widely
and scalability.1 Generally, ball milling or grinding in a mortar used. However, these are applied primarily for material proces-
and pestle does not allow the control of the temperature. sing, on the tonne scale, and have not been demonstrated for
chemical synthesis.14 Overall, there is a need to be able to
control the temperature of mechanochemical processes as well
aSchool of Chemistry and Chemical Engineering, Queen’s University Belfast,
David Keir Building, Stranmillis Road, Belfast, Northern Ireland, BT9 5AG, UK.
E-mail: d.crawford@qub.ac.uk, s.james@qub.ac.uk
as improve their scalability if mechanochemistry is to become
widely applicable as a chemical manufacturing technique.
Recently, twin screw extrusion (TSE), previously used exten-
sively in the food, polymer and pharmaceutical industries,15
has shown great potential for the efficient, continuous
mechanochemical synthesis of chemicals, including co-crystals,2
bUniversity of Limerick, Bernal Institute, Department of Chemistry & Environmental
Science, Limerick, Ireland
†Electronic supplementary information (ESI) available. See DOI: 10.1039/
c6gc03413f
This journal is © The Royal Society of Chemistry 2017
Green Chem.

View Article Online
Paper
Green Chemistry
metal organic frameworks (MOFs)3 and deep eutectic sol- Space Time Yields (STYs) compared to conventional solvent-
vents.4 TSE involves the conveyance of the material through a based processes. The STY (kg of product obtained per m3 of
confined space (barrel) by the rotation of a pair of intermesh- reactor volume per day) is a parameter which indicates the
ing, co- or counter-rotating screws (Fig. 1).16 The use of overall efficiency of a process. For the synthesis of MOFs, STYs
modular screws allows the screw configuration to be custo- were up to three orders of magnitude greater than those for
mised to meet the specific needs of a given process. Modifying the reported conventional syntheses3 and for the preparation
this parameter allows the control of the total shear and com- of DESs, four orders of magnitude greater than that for the
pressive forces experienced by the material as it moves along standard batch heating procedure.4
the extruder barrel16 and thus the screw design can potentially
Along with increasing interest in the formation of inorganic
be manipulated to instigate and optimise a given chemical and coordination compounds by mechanochemical means,
reaction. Furthermore, the temperature, screw speed and feed there has also been significant interest in carrying out organic
rate of the starting reagents can all be varied, with the latter synthesis this way, and a diverse range of organic reactions
having a direct effect on the overall compression force experi- have been demonstrated, mainly through ball milling.12 In par-
enced by the material.16
ticular, several successful mechanochemical condensation
TSE has been used extensively for the reactive extrusion of reactions have been reported, including the Knoevenagel con-
polymers, including the functionalisation of polymers, densation and aldol reaction.8–10,18 Other examples of organic
through living or polycondensation polymerisation for reactions performed mechanochemically include the Wittig
example.17 The first non-polymer use of the TSE technique in reaction, Diels–Alder, Click and Suzuki reactions, some of
chemical synthesis was in hot melt extrusion to form co- which usually require quite stringent conditions in their con-
crystals2 in which Amgen report the synthesis of AMG-517 – ventional solution-based synthesis such as the use of dry sol-
sorbic acid co-crystal via TSE.2 These authors also highlight vents and an inert atmosphere. However, there are several
additional advantages of employing TSE in addition to the examples where ball milling has been performed successfully
benefits related to the process efficiency. In particular, they without such rigorous conditions.19
report that the cocrystals obtained exhibit enhanced properties
Some research into the scaling up of mechanochemical
(compared to solution grown cocrystals), including a higher organic reactions has been carried out, particularly by Stolle
surface area (2.9 m2 g−1 vs. 0.9 m2 g−1), thereby eliminating et al. who investigated the Knoevenagel condensation between
the need for post-synthesis processing involving milling.
vanillin and barbituric acid, scaling up from mixer to plane-
More recently we have demonstrated the synthesis of Metal tary ball mills.9 Stolle considered factors such as milling fre-
Organic Frameworks (MOFs),3 Deep Eutectic Solvents (DESs)4 quency and time, as well as the vessel size and shape, ball dia-
and discrete metal complexes by TSE.3 In most cases (with meter, weight and reagent fill level, finding that they all had
exception of HKUST-1 and [Ni(salen)]3), the reactions were important effects on the reaction process. The maximum scale
carried out without added solvents and gave materials of of this study was 400 g (using four 100 g grinding stations).9
similar or superior quality to those prepared by conventional
As the formation of co-crystals and inorganic materials has
means. Notably, the fast reaction times involved in TSE been scaled up to a continuous manufacturing process through
(0.5–4 minutes) meant that the problematic thermal degra- the use of TSE, it is interesting now to investigate organic
dation typically observed in the conventional 12 h preparation synthesis by this technique. Here, we report four condensation
of the choline chloride : D-fructose DES, as a consequence of reactions successfully carried out on a large scale by TSE,
fructose caramelisation, was avoided.4 The elimination or specifically the Knoevenagel condensation, imine formation,
reduction of the solvent used not only provided more sustain- the aldol reaction and the Michael addition. These processes
able synthetic processes but also dramatically increased the could be optimised such that each product was synthesised
without any added solvent and in many cases, analytically pure
products were obtained without any post process purification.
Results and discussion
Knoevenagel condensation
Kaupp et al. first investigated the Knoevenagel condensation
between vanillin and barbituric acid in the solid state,10 and
Stolle et al. successfully highlighted the potential of this reac-
tion to be carried out on a larger scale mechanochemically, via
planetary ball milling.8,9 We therefore chose this as a bench-
mark organic reaction to be studied by TSE, and subsequently
extended it to include the reaction of barbituric acid with
two other aldehydes, veratraldehyde and 5-bromovanillin
(Scheme 1). A comparison between the three aldehydes
Fig. 1 (a) Twin screw extruder and (b) intermeshing co-rotating screws
encased in the extruder barrel.
Green Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Green Chemistry
Paper
Fig. 2 Graph outlining the conversion as a variance of time, for the
three Knoevenagel condensations with barbituric acid in anhydrous d6-
DMSO.
sions. As the main aim of this work did not involve investi-
gating the detailed kinetics of each reaction, the corres-
ponding reaction rates in d6-DMSO were not calculated.
However, the equilibrium conversions for each of the three
reactions and the time required to reach equilibrium were
taken as indications as to how readily these reactions pro-
ceeded in solution.
Scheme 1 Reaction of three different aldehydes with barbituric acid in
the solid state, by Twin Screw Extrusion (TSE).
A comparison of the conversion vs. time curves established
that all three reactions reached equilibrium at around day 50.
However, a trend was established upon comparison of these
equilibrium conversions, which indicated that the extent of
the reaction is dependent on the electronic substituents
present on the aromatic ring, as could be expected.20a The
reaction is favoured by electron-donating substituents para to
the aldehyde group, as is the case with vanillin. When a
methoxy group (as in veratraldehyde) is in the para position,
which is considered to be less electron donating than a
hydroxyl group,20b the conversion at equilibrium is ca. 20%
less. Similarly, comparing vanillin and 5-bromovanillin, the
addition of an electron withdrawing bromine group to the aro-
matic ring decreases the conversion to the product.
In regard to the ball mill reactions between barbituric acid
and each of the three aldehydes, it was seen that two out of
three reactions went to completion. The reaction of vanillin and
barbituric acid was complete after 60 minutes at 25 Hz whilst
the reaction involving veratraldehyde required 90 minutes to
reach 100%. However, at 90 minutes (maximum milling time
available) the conversion observed for the reaction between
5-bromovanillin and barbituric acid was only 77.3% (Table 2).
Hutchings et al. reported that for the milling-induced
Knoevenagel condensation, a feedback mechanism based on
provides insight as to whether reactant melting points (Table 1)
are important factors in this type of synthesis (i.e. the aldehyde
with the lowest melting point could be expected to react most
readily) or whether electronic substituent effects dominate as
in solution (i.e. the aldehyde with electron donating substitu-
ents would be expected to react most readily). Furthermore,
the Knoevenagel reaction between vanillin and malononitrile
was also investigated to determine the effect of replacing the
activated methylene reactant, barbituric acid, with a com-
pound of substantially lower melting point.
For comparison with the TSE results, all three reactions
were initially carried out by ball milling as well as in anhy-
drous d6-DMSO, and were monitored by 1H NMR spectroscopy,
allowing any differences in the reactivities of the aldehydes in
the solid state and in solution to be determined. Interestingly,
it was found that a similar trend in reactivity was observed for
both of these methods, with the reaction between vanillin and
barbituric acid occurring most rapidly, whilst the reaction
involving 5-bromovanillin was the slowest.
With respect to the solution based reactions, the conversion
vs. time graphs (Fig. 2) show that the reactions are slow (tens
of days) and do not actually go to completion. We believe that
each reaction reaches equilibrium at these maximum conver-
Table 2 Conversion to the product via ball milling, as determined by 1H
Table 1 Melting points of the various starting reagents
NMR spectroscopy
Reagent
Melting point (°C)
Conversion to product (%)
Veratraldehyde
Vanillin
5-Bromovanillin
Barbituric acid
Malononitrile
40–43
81–83
162–166
245
Time (min)
1a
1b
1c
30
60
90
66.6
99.3
—
12.1
90.50
98.61
54.4
76.8
77.3
32
This journal is © The Royal Society of Chemistry 2017
Green Chem.

View Article Online
Paper
Green Chemistry
both the chemical and physical properties of the system was in
operation, with which there was a significant dependency on the
temperature reached inside the ball mill as well as the changes
in the rheology of the reaction mixture during milling.21
Therefore, restarting the ball milling process to prolong the reac-
tion time would disrupt the temperature and possibly the rhe-
ology of the mixture, leading to inaccurate results.
Furthermore, pseudo-sigmoidal kinetics are observed for
each of these ball milling reactions in accordance with the
unpublished results by Hutchings et al. (Fig. 3), suggesting
that the same feedback mechanism also operates in
the Knoevenagel condensation between veratraldehyde or
5-bromovanillin and barbituric acid.21
Fig. 4 Mixture of vanillin and barbituric acid before extrusion (left) and
the extruded product, 1a (right).
The results observed here are surprising as it was expected
that the ball milling reactions would show dependency on the
melting points of each aldehyde, with veratraldehyde of the
lowest melting point reacting the fastest. This expectation was
based on previous ball milling investigations carried out by
James and Pichon.22 However, this is not the case as vanillin
reacts the fastest. If we consider the time required for the reac-
tions to go to completion, a trend of OH < OMe < Br is appar-
ent, and one would immediately assume that the reactions
occur based on the effect of the electronic substituents.
However, considering the unconventional shapes of the con-
version curves, it can be presumed that the mechanism by
which these reactions are proceeding in the solid state is not
straightforward and requires further investigation which is
beyond the scope of this work.
With our main goal involving the scale-up of these reac-
tions into a continuous, high yielding process, we next investi-
gated the Knoevenagel condensation via twin screw extrusion
and showed that 1a–c could be synthesised successfully via
this technique. Colour changes were seen upon the reaction of
each aldehyde with barbituric acid, i.e. from white to orange
(Fig. 4) which gave an indication that the reaction was proceed-
ing. Characterisation was carried out by 1H NMR spectroscopy,
PXRD and elemental analysis, which confirmed that in each
optimised reaction, an analytically pure product was obtained
as described in detail (Fig. 5 and 6).
Fig. 3 Graph showing the conversion versus time for the three
Knoevenagel condensations with barbituric acid by ball milling.
Fig. 5 1H NMR spectra of 1a–c in d6-DMSO.
Green Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Green Chemistry
Paper
Table 3 Feed rate, throughput rate and space time yield for each extru-
sion process in the synthesis of 1a–c
Feed rate
Throughput rate
Space time yield
Product
(kg h−1
)
(kg h−1
)
(kg m−3 day−1
)
1a
1b
1c
0.599
0.254
0.433
0.520
0.228
0.398
258 385
113 291
197 942
However, the screw speed of 55 rpm resulted in optimum
mixing (indicated by complete conversions) with a residence
time of ca. 2 minutes. At 30 rpm, the residence time was
slightly longer at just over 2 minutes, however the highest con-
versions obtained were then reduced to 95–98%. Therefore,
the better mixing at 55 rpm may have improved the contact
Fig. 6 PXRD patterns of products 1a–c prepared by Twin Screw
Extrusion (TSE).
TSE allows precise temperature control across the different between the reactive surfaces of the materials.
heating zones making up the extruder barrel. Therefore,
Table 3 reports the feed rates optimised for each reaction
experiments were carried out between room temperature and when the extrusion processes were carried out at 160 °C and
160 °C. This range allowed the investigation of the reactions 55 rpm, with the highest feed rate of 0.5 kg h−1 being found
below reactant melting points or above the melting point of for the reaction between vanillin and barbituric acid. The lim-
the aldehyde. A standard screw speed of 55 rpm was initially iting factor preventing greater feed rates was the high levels of
employed as it provides efficient mixing as well as a reasonable torque recorded as a result of over-filling the extruder barrel.
residence time (i.e. the time required to travel along the length However, the level of torque with which an extruder can with-
of the extruder) of ca. 2 minutes.
stand increases with the size of the extruder. We employed a
At room temperature, the starting materials were recovered small extruder (screw diameter 12 mm, 40 : 1 L/D) to fit a stan-
with no trace of product, even after several passes through the dard fume cupboard.
extruder. Above 40 °C partial conversion to products was
After each process was optimised to give complete conver-
observed. However, for all three reactions, 100% conversion sion, its efficiency was assessed by considering the throughput
was achieved at 160 °C, which is significantly higher than the rate and the space time yields (Table 3). Twin screw extrusion
melting points of veratraldehyde and vanillin, but similar to has been shown to result in high STYs, e.g. in the synthesis of
that of 5-bromovanillin.
MOFs,3 as a result of the low free volume present in the extruder
We then optimised the screw speed. Increasing the screw barrel and removing or reducing the amount of solvents. This is
speed not only results in greater mixing of the material and a again the case in this work with all three Knoevenagel conden-
greater shear and mechanical energy being applied to the sations having high STYs with the highest being for the reaction
material, but also reduces the residence time, which may then between vanillin and barbituric acid (>250 000 kg m−3 day−1).
be too short for the complete reaction. Longer residence times
Again, as with the ball mill reactions, the ease of the reac-
are achievable by altering the screw configuration, as will be tion did not correlate with the reactant melting points.
discussed below for the Michael addition. Fig. 7 shows the Therefore, DSC analysis was conducted on stoichiometric mix-
screw configuration employed for the Knoevenagel conden- tures of each aldehyde with barbituric acid at various heating
sations, which consisted largely of conveying sections (with rates (for full details see the ESI†) to gain greater insights into
the channel depth decreasing from one conveying section to the effects of heating on these mixtures.
the next) and kneading segments containing combinations of
30°, 60° and 90° orientations of the segments, with greater whilst the reactants were being analysed in the open DSC pan,
angles applying the greater shear to the material.
indicated by the presence of an endothermic peak represent-
It was observed that a reaction was initiated in each case
Screw speeds of 30–250 rpm were investigated, with speeds ing the product. As an example, Fig. 8 shows a DSC curve of a
>200 rpm resulting in residence times of <30 seconds. mixture of vanillin and barbituric acid with a peak at
ca. 300 °C which corresponds to the melting of the corres-
ponding product.
In addition, the peaks are present representing the melting
of the starting materials at ca. 80 °C for vanillin and ca. 260 °C
for barbituric acid. There appears to be an initial shoulder on
the endothermic peak for the melting of vanillin. We believe
that this peak shape is due to a combination of surface and
bulk melting of the aldehyde.
The most revealing part of the DSC curve is the region from
120 to 160 °C, which shows two exothermic peaks and one
Fig. 7 Screw
condensation.
configuration
employed
in
the
Knoevenagel
This journal is © The Royal Society of Chemistry 2017
Green Chem.

View Article Online
Paper
Green Chemistry
Fig. 8 DSC curve of the vanillin and barbituric acid at a heating rate of
5 °C min−1
.
endothermic peak. This indicates that more than one process
occurs. The endothermic peak can be assigned to the
Knoevenagel reaction occurring, as supported by the litera-
ture.23 Several possibilities such as solidification of the formed
product, polymerisation of the starting material or hydration
of barbituric acid – to form its known dihydrate24 – were con-
sidered to account for the exothermic peaks, but none ade-
quately explained the results observed, both in the DSC curve
and in the extrusion reactions.
Nevertheless, as these reactions are each between a solid
and a molten reactant, it was considered that the wettability of
the reagents may be key to explaining why conditions of
160 °C and 55 rpm were essential for complete conversion.
Wettability experiments were conducted with vanillin or vera-
traldehyde, each in a melt phase, dropped onto a compressed
disc of barbituric acid (Fig. 9). The contact angles were
measured with time at the respective melting points of each
aldehyde as well as at 155 °C to mimic the extrusion and DSC
processes. Unfortunately, as 5-bromovanillin does not melt
until 164 °C, it was not possible to obtain similar data for this
material.
Fig. 10 Graph of contact angle versus time between barbituric acid and
vanillin and between barbituric acid and veratraldehyde at each alde-
hyde’s respective melting points and at 155 °C.
Fig. 10 shows that at their respective melting points, the
contact angles of both vanillin (54.9°) and veratraldehyde (68°)
decrease to ca. 11° within 3 seconds. In fact, within the first
seconds of contact with barbituric acid, the contact angle of
vanillin decreases by ca. 35°, and veratraldehyde undergoes a
more dramatic decrease of ca. 50°. More crucially, for experi-
ments conducted at 155 °C the initial contact angles are very
similar for both aldehydes (34° for veratraldehyde and 31° for
vanillin), and they decrease to similar angles (8° for veratr-
aldehyde and 11° for vanillin), within the same time period. In
addition, Fig. 11 highlights the time required for each alde-
hyde to reach a contact angle of 20° (below which reagents are
Fig. 11 Bar chart representing the wetting time required for vanillin and
veratraldehyde to reach an optimum contact angle of 20°, at both their
respective melting temperatures as well as 155 °C.
Fig. 9 Molten aldehyde dropped onto a compressed disc of barbituric
acid.
Green Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Green Chemistry
Paper
considered to have high wettability), at these various tempera-
tures. At each aldehyde’s respective melting point, vanillin
takes
a greater time (1.2 seconds) than veratraldehyde
(0.8 seconds) to reach 20°. However, at 155 °C the time
required to reach 20° is exactly the same for both vanillin and
veratraldehyde (0.5 seconds).
Therefore, as at 155 °C, the contact angles measured for
each aldehyde and the time required to reach a low angle of
20° were the same, we suggest that for the Knoevenagel con-
densation under these conditions it is the wettability of the
reagents that is vital for the reaction to proceed to completion.
Decreasing the temperature of the reaction or even the time in
the extruder would hinder the wetting of barbituric acid by the
aldehyde, thereby slowing the reaction.
The Knoevenagel reaction between vanillin and malono-
nitrile was also carried out by TSE. Malononitrile, an alterna-
tive activated methylene compound, has a lower melting point
(32 °C) than barbituric acid (164 °C), allowing us to investigate
the effect of this aspect on the previous reactions. Kaupp
reported this solvent-free reaction in the presence of Na2CO3
(0.1 mol. eq.) (Scheme 2) by ball milling.10 Unfortunately, we
could not reproduce this reaction in the ball mill, obtaining
instead a mixture of unidentified products together with the
starting materials. However, in our hands the reaction was suc-
cessful by TSE as described below.
Fig. 12 1H NMR spectrum of 2 in d6-DMSO.
Preliminary extrusion experiments were carried out in the
absence of a catalyst at barrel temperatures ranging from
ambient temperature to 160 °C. However, malononitrile was
suspected to have polymerised upon heating (a common
phenomenon with this reagent)25 and no product was
observed. Therefore, Na2CO3 (0.1 mol. eq.) was employed
during the subsequent experiments. 1H NMR spectroscopy
and PXRD analysis (Fig. 12 and 13) indicated that complete
conversion to product, 2, occurred at a screw speed of 55 rpm
(as before), but at a lower temperature (120 °C) than that
required for the reactions with barbituric acid (160 °C). A
maximum feed rate of 0.21 kg h−1 was achieved, with a
throughput rate of 0.196 kg h−1, corresponding to a high STY
Fig. 13 PXRD patterns of 2 prepared in solution, and by twin screw
extrusion.
ture (120 °C) was much greater than the melting points of
either reagent.
Finally, to explore whether extrusion can be applied to mix-
tures of solid and liquid reagents, we next investigated the
reaction between solid vanillin and liquid ethyl cyanoacetate
in the presence of Na2CO3 (10 mol%) (Scheme 3). A syringe
pump was used to feed ethyl cyanoacetate into the extruder
and a volumetric feeder supplied the solid mixture of vanillin
and Na2CO3 (Fig. 14). The feed rates were optimised as
1.74 ml min−1 and 2.9 g min−1, respectively, resulting in a
slightly longer residence time of 6–7 minutes (compared to the
of 98 × 103 kg m−3 day−1
.
Inclusion of a catalyst means that a direct comparison of
the required process temperatures for the Knoevenagel con-
densations involving barbituric acid versus malononitrile
cannot be made. However, one key observation from the latter
reaction is that it is not simply the formation of a melt phase
that results in a complete conversion, as the process tempera-
Scheme 2 Knoevenagel condensation carried out by TSE, 55 rpm,
120 °C.
Scheme 3 Knoevenagel condensation reaction to prepare 3.
This journal is © The Royal Society of Chemistry 2017
Green Chem.

View Article Online
Paper
Green Chemistry
ponds to a Space Time Yield (STY) of 55 650 kg m−3 day−1. The
catalyst, Na2CO3, was not removed from the extrudate before
analysis (to avoid continuation of the reaction during purifi-
cation). Allowing for 0.1 mol. eq. Na2CO3 in the extrudate, the
pure product was obtained as indicated by elemental analysis.
It can be noted that in solution morpholine (more difficult to
remove than Na2CO3) is conventionally used to catalyse this
reaction.
To summarise regarding the Knoevenagel condensations,
the above study represents the first step in scaling up solvent-
free organic synthesis by extrusion. This technique was found
Fig. 14 Twin screw extruder with a volumetric feeder (A) (feeding vanil- to be applicable to a range of reactions, including those
lin and Na2CO3), and a syringe pump (B) (feeding ethyl cyanoacetate).
between a solid and a liquid as well as reactions involving
reagents of varying melting points. In most examples, post syn-
thetic purification was not required, and in these cases, TSE
previous Knoevenagel condensations with residence times of
ca. 2 minutes). The product of this reaction has been investi-
gated as a potential photostabiliser in sunscreen products.26
Extrusion experiments were carried out to optimise the
temperature and screw speed required to achieve 100% conver-
sion to a crystalline product (Fig. 15 and 16). A complete reac-
tion was achieved at 160 °C with a screw speed of 55 rpm. A
throughput rate of 0.11 kg h−1 was achieved, which corres-
was shown to produce analytically pure compounds with high
space time yields. Furthermore, it was demonstrated that the
conditions for effective reactions involving barbituric acid cor-
responded to conditions required for effective wetting by the
aldehyde.
Imine formation
Solvent-free C–N bond formation has been reported extensively
in the literature,12 including the formation of unsymmetrical
aromatic diimines as well the synthesis of salenH2 from two
liquids (salicylaldehyde and ethylene diamine),27 both reac-
tions being conducted by ball milling. Herein, we report the
formation of a C–N bond by TSE as demonstrated through the
synthesis of a diimine 4 (Scheme 4). The solution reaction
involves refluxing 4,4′-oxydianiline with 2 equivalents of ortho-
vanillin in methanol for several hours.28
The imine is formed within 30 minutes of ball milling at
25 Hz, without the addition of a solvent. The reaction was
Fig. 15 1H NMR spectrum of 3 in d6-DMSO.
Scheme 4 Mechanochemical imine formation (4) by ball milling at 25
Fig. 16 PXRD pattern of 3 prepared by twin screw extrusion.
Hz for 30 minutes and by TSE at 55 rpm, 120 °C.
Green Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Green Chemistry
Paper
Table 4 Variation of temperature and feed rate in the extrusion of 4,4’-
dioxyaniline with o-vanillin
Temperature
(°C)
Screw speed
(rpm)
Feed rate
Conversion
(%)
(g min−1
)
a
25
40
80
80
80
80
100
120
55
55
55
55
55
55
55
55
—
0
80
99
0
45
70
91
100
a
—
a
—
0.45
0.73
0.79
0.79
0.79
Fig. 17 Orange solid (4) produced from the twin screw extrusion of
4,4’-oxydianiline and 2 equivalents of ortho-vanillin.
a Reagents were fed manually.
subsequently scaled from 0.5 g per batch (by ball milling) to
0.03 kg h−1 via twin screw extrusion, producing the desired
Schiff base as a bright orange powder (Fig. 17). Again, the
temperature, screw speed and feed rate were optimised indivi-
dually for this reaction and an analytically pure product (as
indicated by elemental analysis) was obtained by extruding at
120 °C, 55 rpm with a feed rate of 0.79 g min−1 (Fig. 18 and
19). No post synthetic purification was needed.
Only a partial reaction (28% conversion) occurred at room
temperature, as determined by 1H NMR spectroscopy. At room
temperature, the screw speed was increased to 100, 150 and
250 rpm to encourage greater conversion, however this actually
resulted in lower conversions as a result of the shorter resi-
dence times.
Table 4 shows the temperatures and feed rates investigated
as well as the conversions to 4 (as determined by 1H NMR
spectroscopy). Moderate temperatures of 40–80 °C resulted in
significant conversions to the product of up to 99%. However,
it was quickly identified that increasing the temperature
further allowed a greater feed rate and 100% conversion. This
was considered to be more advantageous than employing a
lower temperature, and consequently at 120 °C and 55 rpm
(with a residence time of ca. 30–60 seconds), a feed rate of
0.79 g min−1 was achieved. This is almost double the feed rate
of the experiment carried out at 80 °C.
Therefore, the formation of a C–N bond has been carried out
as a solvent-free continuous process, made possible by TSE. The
analytically pure product (as determined by elemental analysis)
is obtained after ca. 2 minutes without any post-synthetic
workup. Furthermore, a throughput rate of 0.03 kg h−1 was
achieved resulting in a high STY of 14 900 kg m−3 day−1
.
Fig. 18 1H NMR spectrum of Schiff base 4 in d6-DMSO.
Michael addition
Michael addition is another subset of organic reactions that
has been reported to proceed efficiently by ball milling, for
example, the reaction between vanillin and dimedone as
reported by Kaupp et al.10 Herein, we report a related reaction
between veratraldehyde and dimedone (2 equivalents) by TSE
(Scheme 5).
Unfortunately, initial ball mill experiments were unsuccess-
ful. Milling of the starting materials at 25 Hz for 60 minutes
resulted in a low conversion to the product of only 7.5%.
Experiments employing liquid assisted grinding (with CHCl3
or H2O) and/or catalysts (CaCl2 and SmBr3·6H2O, reported to
work in the solution reactions)29,30 were also unsuccessful. In
fact the use of CaCl2 was observed to hinder the ball mill reac-
tion, resulting in <1.5% conversion to the product.
However, the control over temperature and screw configur-
ation offered by extrusion methods encouraged us to attempt
this reaction in the extruder.
Fig. 19 PXRD patterns of the Schiff base prepared in solution, by ball
milling and twin screw extrusion.
This journal is © The Royal Society of Chemistry 2017
Green Chem.

View Article Online
Paper
Green Chemistry
achieved reproducibly. It must be noted that no reaction had
occurred after mixing in the planetary ball mill, as determined
by 1H NMR spectroscopy.
Although there was 100% conversion of starting materials,
more than one product was observed to form in this reaction.
The dihydroxyl compound 5 was the major product, and un-
identified byproducts were present in trace amounts. The
same mixture of products, in the same ratio, was observed to
form regardless of the synthetic technique or catalyst used, i.e.
the same result was obtained from both solvent-based and
mechanochemical experiments. Therefore, the formation of
these byproducts in the extruder was not considered to be a
significant drawback compared to the solution-based method.
A sample of the obtained extrudate was purified by column
chromatography (eluent system CH2Cl2 : EtOH 9 : 1) to produce
Scheme 5 Michael addition carried out mechanochemically to form 5.
At room temperature and screw speeds of 30–250 rpm, no
reaction was observed to occur with a white solid consisting of
the starting materials being retrieved. However, a slight temp-
erature increase to 40 °C resulted in a significant change in
the properties of the extrudate, producing a golden yellow
liquid which rapidly formed glass on cooling. However, 1H
NMR spectroscopy indicated that the extrudate still consisted
mainly of starting reagents (<5% conversion to the product).
Increasing the temperature further did increase the conver-
sion to the product, e.g. at 150 °C and 55 rpm a conversion of
80% was achieved. Increasing the temperature above 160 °C
resulted in a brown viscous liquid, which is believed to result
from the thermal degradation of one of the reagents.
Therefore, as degradation was observed at high temperatures,
and increasing screw speed resulted in insufficient residence
times, the screw configuration was then changed to determine
if this could drive the reaction to completion.
The alternative screw configuration consisted of conveying
and kneading segments as before, but with the addition of
reverse conveying segments (Fig. 20). These segments retard
the flow of the material through the extruder increasing the
residence time.16 They were strategically positioned after each
kneading section, to hold the material in the high shear
kneading sections for a prolonged period of time.
1
an analytically pure white solid (5) as determined by H NMR
spectroscopy, PXRD and elemental analysis (Fig. 21 and 22).
As expected, the residence time of the optimised process
was longer at ca. 12 minutes, due to the presence of reverse
conveying segments (compared with 2 minutes for the stan-
dard screw configuration in the Knoevenagel condensations).
An optimised feed rate of 0.12 kg h−1 resulted in an overall
throughput rate of 0.07 kg h−1. A space time yield of 35 × 103
kg m−3 day−1 was determined for these conditions.
When this reaction is conducted in solution, catalysts such
as CaCl2 or SmBr3·6H2O are used. The catalyst and the temp-
erature determine the final product obtained from the reac-
tion.29,30 The dihydroxyl product 5 obtained in this work is
typically formed at room temperature in solution. However, at
reflux an alternative xanthene product 5′ is isolated instead
(Fig. 23). Therefore, not only have we demonstrated that TSE
can be used to carry out Michael additions, but also that extru-
sion can give a different chemoselectivity from that observed
in solution.
Aldol condensation
Using this configuration, complete conversion to the
product was achieved at 120 °C and 200 rpm to produce a
beige/yellow solid. However, the repetition of this result proved
difficult with conversions to the product being observed there-
after between 97 and 98%. After some consideration, the
problem was identified to be inefficient premixing before
extrusion. The reagents were of similar particle sizes
(ca. 300 µm) and consistencies but it was only when they were
premixed in a planetary ball mill in the absence of any grind-
ing media (170 rpm, 5 minutes), followed by extrusion at
120 °C, 200 rpm, that 100% conversion to the product was
Aldol condensations have been investigated extensively by ball
milling, and demonstrate chemoselectivity for one product in
Fig. 20 Alternative screw configuration employed, containing reverse
conveying segments that retard the flow of material.
Fig. 21 1H NMR spectrum of 5 in CD3OD.
Green Chem.
This journal is © The Royal Society of Chemistry 2017

View Article Online
Green Chemistry
Paper
Fig. 22 PXRD patterns of 5 prepared in solution, by ball milling and
TSE.
Fig. 24 1H NMR spectrum of the 1,3 indandione, 6 in d6-DMSO.
Fig. 23 Possible products from the Michael addition.
Fig. 25 PXRD patterns of the ball mill and extrusion products of the
aldol reaction.
many cases. Kaupp et al. reported the reaction between ninhy-
drin and dimedone by ball milling, to produce a substituted
1,3 indandione 6 (Scheme 6).31 We were able to reproduce this
result by ball milling and therefore subsequently investigated
the reaction by TSE. Initial results showed that the reaction
would not go to completion without the presence of Na2CO3
(0.1 mol. eq.), even when an alternative screw configuration
was employed to prolong the residence time (as with the
Michael addition).
At 75 °C and 55 rpm, 80% conversion to the product was
achieved, but increasing the temperature to 100 °C gave 100%
conversion to the product. At this temperature, screw speeds of
30–200 rpm were investigated, and 100% conversion to the
product was observed at 55 rpm (Fig. 24 and 25).
long (12 minutes), in comparison to ca. 2 minutes (as in the
Knoevenagel condensations). Furthermore,
a
significant
increase in torque was also observed. This is believed to result
from a change in the rheological properties of the material
during the course of the reaction. Due to this high torque, the
feed rate was kept low at 0.8 g min−1, to prevent an obstruction
occurring along the barrel, disrupting the continuous process.
A throughput rate of 63.6 g h−1 was obtained, resulting in a
Space Time Yield (STY) of 32 × 103 kg m−3 day−1
.
The reported conventional synthesis involves refluxing in
water overnight.32 However, we found that refluxing in water
for 3 days, even in the presence of 0.1 mol. eq. Na2CO3 did not
result in a complete reaction (78% conversion to the product).
Ultrasound irradiation has also been employed to carry out
this reaction,33 however, it is still very difficult to scale up, sup-
porting the use of extrusion for the large scale, continuous
synthesis of a substituted 1,3 indandione 6.
A standard screw configuration was employed, and it was
observed that the residence time in this case was unusually
Conclusions
In conclusion, we have demonstrated through a variety of con-
densation reactions that organic synthesis can be conducted
Scheme 6 Aldol condensation carried out by TSE at 55 rpm, 100 °C to
form 6.
as
a
large scale, continuous process under solvent-free
This journal is © The Royal Society of Chemistry 2017
Green Chem.

View Article Online
Paper
Green Chemistry
conditions. In most cases, post process purification was not 10 G. Kaupp, M. Reza Naimi-Jamal and J. Schmeyers,
required and high throughput rates (ca. 0.5 kg h−1) and space
Tetrahedron, 2003, 59, 3753.
time yields (up to 260 × 103 kg m−3 day−1) were achieved. The 11 V. P. Balema, J. W. Wiench, M. Pruski and V. K. Pecharsky,
physical mechanism by which some of the Knoevenagel con- J. Am. Chem. Soc., 2002, 124, 6244.
densations took place was studied and identified as involving 12 G.-W. Wang, Chem. Soc. Rev., 2013, 42, 7668.
wetting of barbituric acid by the liquid aldehyde as a critical 13 (a) L. Takacs and J. S. McHenry, J. Mater. Sci., 2006, 41,
step. Furthermore, the first example of a reaction (Michael
addition) that shows a different chemoselectivity under TSE
compared to that in solution has also been presented.
5246; (b) M. Poux, P. Cognet and C. Gourdon, Green Process
Engineering: From Concepts to Industrial Applications, CRC
Press, Taylor and Francis Group, Boca Raton, 2010;
(c) K. Uzarevič, V. Štrukil, C. Mottillo, P. A. Julien,
A. Puskarič, T. Frisčič and I. Halasz, Cryst. Growth Des.,
2016, 16, 2342; (d) X. Ma, W. Yuan, S. E. J. Bell and
S. L. James, Chem. Commun., 2014, 50, 1585.
Acknowledgements
14 G. Le Caër, P. Delcroix, S. Bégin-Colin and T. Ziller,
Hyperfine Interact., 2002, 141/142, 63.
We thank EPSRC for funding (EP/L019655/1).
15 D. E. Crawford and J. Casaban, Adv. Mater., 2016, 28, 5747.
16 H. F. Giles, J. R. Wagner and E. M. Mount, Extrusion: The
definitive processing guide and handbook, William Andrew,
Elsevier, Amsterdam, 2014.
Notes and references
1 (a) G. A. Bowmaker, Chem. Commun., 2013, 49, 334; 17 M. E. Hyun and S. C. Kim, Polym. Eng. Sci., 1988, 28, 743.
(b) G.-W. Wang, Chem. Soc. Rev., 2013, 42, 7535; (c) P. Baláž, 18 C. L. Raston and J. L. Scott, Green Chem., 2000, 2, 49.
M. Achimovičová, M. Baláž, P. Billik, Z. Cherkezova- 19 F. Ravalico, S. L. James and J. S. Vyle, Green Chem., 2011,
Zheleva, J. M. Criado, F. Delogu, E. Dutková, E. Gaffet,
13, 1778.
F. J. Gotor, R. Kumar, I. Mitov, T. Rojac, M. Senna, 20 (a) J. Clayden, N. Greeves, S. G. Warren and P. Wothers,
A. Streletskii and K. Wieczorek-Ciurowa, Chem. Soc. Rev.,
2013, 42, 7571; (d) V. Šepelák, A. Düvel, M. Wilkening,
K.-D. Becker and P. Heitjans, Chem. Soc. Rev., 2013, 42,
Organic Chemistry, Oxford University Press, 2nd edn, 2012;
(b) J. E. Leffler and E. Grunwald, Rates and Equilibria of
Organic Reactions, Wiley, 1963.
7507; (e) D. Braga, L. Maini and F. Grepioni, Chem. Soc. 21 B. Hutchings, Feedback kinetics in a mechanochemical reac-
Rev., 2013, 42, 7638; (f) K. Ralphs, C. Hardacre and tion, Masters, Queen’s University Belfast, 2015.
S. L. James, Chem. Soc. Rev., 2013, 42, 7701; 22 A. Pichon and S. L. James, CrystEngComm, 2008, 10,
(g) E. Boldyreva, Chem. Soc. Rev., 2013, 42, 7719; 1839.
(h) B. Rodriguez, A. Bruckmann, T. Rantanen and C. Bolm, 23 M. Kolahdoozan, R. J. Kalbasi, Z. S. Shahzeidi and
Adv. Synth. Catal., 2007, 349, 2213; (i) T. Friscic, Chem. Soc.
Rev., 2012, 41, 3493.
2 (a) C. Medina, D. Daurio, K. Nagapudi and F. Alvarez-
F. Zamani, J. Chem., 2013, 2013, 1.
24 G. S. Nichol and W. Clegg, Acta Crystallogr., Sect. B: Struct.
Sci., 2005, 61, 464.
Nunez, J. Pharm. Sci., 2010, 99, 1693; (b) R. S. Dhumal, 25 B. M. Ferrier and N. Campbell, J. Chem. Soc., 1960, 3513.
A. L. Kelly, P. York, P. D. Coates and A. Paradkar, Pharm. 26 R. Chaudhuri, Photostable organic sunscreen composition,
Res., 2010, 27, 2725; (c) D. Daurio, K. Nagapudi, L. Li,
P. Quan and F.-A. Nunez, Faraday Discuss., 2014, 170, 235.
3 D. Crawford, J. Casaban, R. Haydon, N. Giri, T. McNally
and S. L. James, Chem. Sci., 2015, 6, 1645.
4 D. E. Crawford, L. A. Wright, S. L. James and A. P. Abbott,
Chem. Commun., 2016, 52, 4215.
US 20070059258A1, 2004.
27 M. Ferguson, N. Giri, X. Huang, D. Apperley and
S. L. James, Green Chem., 2014, 16, 1374.
28 P. Cucos, F. Tuna, L. Sorace, I. Matei, C. Maxim, S. Shova,
R. Gheorghe, A. Caneschi, M. Hillebrand and M. Andruh,
Inorg. Chem., 2014, 53, 7738.
5 S. L. James, C. J. Adams, C. Bolm, D. Braga, P. Collier, 29 B. Maleki, M. Raei, E. Akbarzadeh, H. Ghasemnejad-Bosra,
T. Friscic, F. Grepioni, K. D. M. Harris, G. Hyett, W. Jones,
A. Krebs, J. Mack, L. Maini, A. G. Orpen, I. P. Parkin,
W. C. Shearouse, J. W. Steed and D. C. Waddell, Chem. Soc. 30 A.
Rev., 2012, 41, 413.
A. Sedrpoushan, S. S. Ashrafi and M. N. Dehdashti, Org.
Prep. Proced. Int., 2016, 48, 62.
Ilangovan,
S.
Muralidharan,
P.
Sakthivel,
S. Malayappasamy, S. Karuppusamy and M. P. Kaushik,
6 D. R. Maurice and T. H. Courtney, Metall. Trans. A, 1990,
Tetrahedron Lett., 2013, 54, 491.
21, 289.
31 G. Kaupp, M. R. Naimi-Jamal and J. Schmeyers, Chem. –
Eur. J., 2002, 8, 594.
7 N. Stock and S. Biswas, Chem. Rev., 2012, 112, 933.
8 A. Stolle, R. Schmidt and K. Jacob, Faraday Discuss., 2014, 32 N. P. Peet, E. W. Huber and J. C. Huffman, J. Heterocycl.
170, 267. Chem., 1995, 32, 33.
9 C. F. Burmeister, A. Stolle, R. Schmidt, K. Jacob, 33 G. Cravotto, A. Demetri, G. M. Nano, G. Palmisano,
S. Breitung-Faes and A. Kwade, Chem. Eng. Technol., 2014,
37, 857.
A. Penoni and S. Tagliapietra, Eur. J. Org. Chem., 2003,
4438.
Green Chem.
This journal is © The Royal Society of Chemistry 2017