Aqueous Kolbe-Schmitt synthesis using resorcinol in a microreactor laboratory rig under high-p,T conditions

Article abstract of DOI:10.1021/op050045q

The aqueous Kolbe-Schmitt synthesis using resorcinol to yield 2,4-dihydroxy benzoic acid was performed in a microreactor rig. This small-scale plant was equipped initially with one capillary reactor and one microstructured cooler only. Later, two upgraded versions were constructed, having in addition a microstructured cooler and a microstructured mixer, respectively. The chemical protocol was significantly varied as compared to standard laboratory operation as described in the literature. Higher temperatures (up to 220°C) and pressures (up to 74 bar) were employed in a facile manner, termed high-p,T processing. In this way, the reaction time could be shortened by orders of magnitude, from about 2 hours to less than one minute, in some cases to some seconds. This resulted in a remarkable increase of the space-time yield by a factor of 440 at best. Productivity was in the L/h range and yielded at best 111 g/h product, corresponding to 4 t/a. Scale-out solutions are indicated. Drawbacks of the microreactor operation were also identified such as high sensitivity to fouling and delicate regulation of the system pressure, leading to partly unstable plant operation. Possibly even a considerable part of the product was rearranged to 2,6-dihydroxybenzoic acid and then thermally decomposed under the harsh reaction conditions. Solutions to overcome or at least diminish these restrictions are envisaged, and in the hope that this may be achieved, a process innovation and business perspective for the high-p,T microreactor processing is depicted.

Full text of DOI:10.1021/op050045q

Organic Process Research & Development 2005, 9, 479−489
Aqueous Kolbe-Schmitt Synthesis Using Resorcinol in a Microreactor
Laboratory Rig under High-p,T Conditions
Volker Hessel,* C. Hofmann, P. Lo¨b, J. Lo¨hndorf, H. Lo¨we, and A. Ziogas
Institut fu¨r Mikrotechnik Mainz GmbH, Carl-Zeiss-Strasse 18-20, 55129 Mainz, Germany
Abstract:
reaction time. The use of higher temperatures has been
The aqueous Kolbe-Schmitt synthesis using resorcinol to yield
2,4-dihydroxy benzoic acid was performed in a microreactor
rig. This small-scale plant was equipped initially with one
capillary reactor and one microstructured cooler only. Later,
two upgraded versions were constructed, having in addition a
microstructured cooler and a microstructured mixer, respec-
tively. The chemical protocol was significantly varied as
compared to standard laboratory operation as described in the
literature. Higher temperatures (up to 220 °C) and pressures
(up to 74 bar) were employed in a facile manner, termed high-
p,T processing. In this way, the reaction time could be shortened
by orders of magnitude, from about 2 hours to less than one
minute, in some cases to some seconds. This resulted in a
remarkable increase of the space-time yield by a factor of 440
at best. Productivity was in the L/h range and yielded at best
111 g/h product, corresponding to 4 t/a. Scale-out solutions are
indicated. Drawbacks of the microreactor operation were also
identified such as high sensitivity to fouling and delicate
regulation of the system pressure, leading to partly unstable
plant operation. Possibly even a considerable part of the product
was rearranged to 2,6-dihydroxybenzoic acid and then ther-
mally decomposed under the harsh reaction conditions. Solu-
tions to overcome or at least diminish these restrictions are
envisaged, and in the hope that this may be achieved, a process
innovation and business perspective for the high-p,T microre-
actor processing is depicted.
reported in a few cases. However, this typically does not
exceed a few 10 °C. Bearing in mind that most of the
protocols take into account the relatively slow mass- and
heat-transfer characteristics of batch reactors (even concern-
ing small laboratory ones), it stands to reason that a
purposefully drastic tailored change in the operating condi-
tions is once required to exploit the full potential of
microreactor technology.
Reactions in batch processes actually often have features
of “domesticated” processing14 (see also ref 15), e.g. as given
by the typical addition of reactants drop-per-drop to exert
control over heat releases of highly exothermic reactions or
the replacement of very harsh reactants by slowly reacting
ones. Actually, batch processing has been established for
centuries,16,17 and the chemistry is made “around the reac-
tor”,17 i.e., adapted; when needed, it is “domesticated”14 rather
than customising the reactor for a chemical process. Now
that expertise on chemical micro process engineering is
growing, it is time to change the chemistry in microstructured
reactors. A list of specific measures has been prepared to
explain what exactly this implies. Among such issues are:14
• operation at unusually high temperatures and pressure
• solvent-free or reactant-rich solution processing
• operation in the explosive regimes
• use of unusually harsh reactants or unstable intermedi-
ates
• combining multiple reactions in a row
In the following, one of the first applications of the first
item is reported, i.e., to change a chemical protocol by
operating at much higher temperatures to reduce reaction
times and, in turn, to increase productivity. Since this
includes setting the pressure much higher to maintain a
single-phase liquid processing, such an approach is termed
high-p,T processing in the following. The Kolbe-Schmitt
synthesis reported in this paper is actually one of the two
1. Need for Adapted Chemical Protocols for Chemical
Micro Process Engineering: Enabling instead of
Subduing Chemistry
Microstructured reactors are these days typically employed
for niche processes where conventional equipment fails or
the performance is much too low.1-15 Typically, in a first
run the batch experimental protocols are mimicked. Often,
the batch protocols are adapted slightly thereafter, the main
difference probably being the notable shortening of the
(6) Ja¨hnisch, K.; Hessel, V.; Lo¨we, H.; Baerns, M. Angew. Chem., Int. Ed.
2004, 43, 406.
(7) Hessel, V.; Lo¨we, H. Chem. Eng. Technol. 2003, 26, 13.
(8) Gavriilidis, A.; Angeli, P.; Cao, E.; Yeong, K. K.; Wan, Y. S. S. Trans.
Inst. Chem. Eng. A 2002, 80, 3.
(9) Jensen, K. F. Chem. Eng. Sci. 2001, 56, 293.
(10) Jensen, K. F. AIChE J. 1999, 45, 2051.
(11) Jensen, K. F. Nature 1998, 393, 735.
(12) Haswell, S. T.; Watts, P. Green Chem. 2003, 5, 240.
(13) Fletcher, P. D. I.; Haswell, S. J.; Pombo-Villar, E.; Warrington, B. H.; Watts,
P.; Wong, S. Y. F.; Zhang, X. Tetrahedron 2002, 58, 4735.
(14) Hessel, V.; Lo¨b, P.; Lo¨we, H. Curr. Org. Chem. 2005, 9, 765.
(15) Hessel, V.; Lo¨b, P.; Lo¨we, H. Chem. Eng. Technol. 2005, 28, 267.
(16) Acricola, G. De Re Metallica; Froben, Episopius: Basel; Vol. XII, p 1556.
(17) Stankiewicz, A. I.; Moulijn, J. A. Chem. Eng. Prog. 2000, 1, 22.
* Corresponding author. Telephone: ++0049-6131-990450. Fax: ++0049-
6131-990305. E-mail: hessel@imm-mainz.de.
(1) Hessel, V.; Hardt, S.; Lo¨we, H. Chemical Micro Process Engineering:
Fundamentals, Modelling and Reactions; Wiley-VCH: Weinheim, 2004.
(2) Hessel, V.; Kolb, G.; Mu¨ller, A.; Lo¨we, H. Chemical Micro Process
Engineering: Processing and Plants; Wiley-VCH: Weinheim, 2005.
(3) Pennemann, H.; Watts, P.; Haswell, S.; Hessel, V.; Lo¨we, H. Org. Process
Res. DeV. 2004, 8, 422.
(4) Pennemann, P.; Hessel, V.; Lo¨we, H. Chem. Eng. Sci. 2004, 59, 4789.
(5) Kolb, G.; Hessel, V. Chem. Eng. J. 2004, 98, 1-38.
10.1021/op050045q CCC: $30.25 © 2005 American Chemical Society
Published on Web 06/07/2005
Vol. 9, No. 4, 2005 / Organic Process Research & Development
•
479

first two reaction examples carried out at IMM to validate
the implementation of the novel concept. The other reaction
was the thermal side-chain monobromination of m-nitro-
toluene which was also presented recently.18
No phenol is lost, and almost a quantitative yield of salicylic
acid is obtained. This now so-called Kolbe-Schmitt method
is still the standard method for the preparation of a wide
variety of aromatic hydroxyl acids.
2. Conventional Kolbe-Schmitt Synthesis and Its
Variants
The Kolbe-Schmitt synthesis is a standard route to
introduce carboxylic moieties into a phenolic core, usually
in ortho position to the hydroxyl group (see a review in ref
19). This reaction route was applied within the well-known
industrial process for the manufacture of the drug Aspirin
of the Bayer Company, used in headache tablets. The
aqueous Kolbe-Schmitt synthesis, which is reported in this
paper, is these days still applied for the synthesis of â- and
γ-rescorcylic acids for uses in dyestuff additives, pharma-
ceuticals, reprographic chemicals, cosmetic preparations,
chelators for resins, and fine organic chemicals.
Many variants of the Kolbe-Schmitt synthesis have been
reported over a period of time, and some of the important
ones shall be briefly given here. The initial work was
performed already about one and a half centuries ago by
Kolbe.20 Salicylic acid was obtained by heating a mixture
of phenol and sodium in the presence of carbon dioxide at
atmospheric pressure.
A further variant, introduced by Marasse, made the
procedure even simpler, while maintaining the performance.23
A mixture of the free phenol and excess anhydrous potassium
carbonate is carbonated under pressure and at elevated
temperatures.
For an industrial application, however, the Marasse variant
is quite expensive due to the costs of the carbonates used.
Carbonations were also made in the presence of solvents.
For the more reactive di- and trihydric phenols such as
resorcinol, pyrogallol, and phloroglucinol monocarbonation
is achieved simply in alkaline aqueous solutions at atmo-
spheric pressure, avoiding the harsh reaction conditions
reported above.24 Resorcinol, for example, is reacted at
ambient pressure in water under reflux conditions with the
weak base KHCO3 to give 2,4-dihydroxybenzoic acid.25,26
Then, a reaction time of approximately 2 h is needed.
The use of organic solvents such as toluene was also
reported for carbonations of phenol at ambient pressure.27
Suspensions formed out of the mixture of reactants and
solvent. This prodcedure is recommended where pressure
equipment is not available.
The conversion of potassium-2-naphthoxide to 2-naphthol
was performed at an industrial scale in the framework of
the Wacker process.28
The mechanism of the Kolbe-Schmitt synthesis was
investigated and an intermediate proposed which predicts that
two sodium phenoxide molecules act as chelating agents for
the carbon dioxide.29,30 In this way, the ortho orientation is
fixed. Recent investigations give more detailed information.
Three intermediates and three transitions are revealed by
applying density functional theory (DFT) modelling.31
This synthesis was not easily scaled out, as there was a
larger request for salicylic acid. Also, the yields on a
laboratory level changed remarkably under seemingly similar
reaction conditions. This was a result of volatilization of the
reactant phenol and the consumption of sodium by reactions
to species other than the salicylic acid anion, e.g., sodium
phenoxide and sodium carbonate. For these reasons the solid
hygroscopic sodium phenoxide, prepared from evaporated
solutions of phenol and sodium hydroxide, was later used
directly as a starting reagent by Kolbe.21
The dried sodium phenoxide was heated in an iron retort
to 180 °C and carbon dioxide passed slowly over the hot
salt. About half of the phenol distilled under these conditions
so that the yield was never larger than 50%.
Schmitt modified the Kolbe procedure by applying
pressure which resulted in greatly improved yields.22 Dry
sodium phenoxide is heated with carbon dioxide at 120-
130 °C for several hours at a pressure of about 80 to 94
bar.
Recent works on the Kolbe-Schmitt synthesis are con-
cerned with the effects of alkali and alkaline earth metals
(23) Marasse, S. German Patent 73,259, 1893; Frdl. 3, 821.
(24) Thiele, J.; Jaeger, K. Ber. 1901, 34, 2840.
(18) Lo¨b, P.; Lo¨we, H.; Hessel, V. J. Fluorine Chem. 2004, 125, 1677.
(19) Lindsey, A. S.; Jeskey, H. Chem. ReV. 1957, 57, 583.
(20) Kolbe, H.; Lautemann, E. Ann. 1860, 113, 125.
(21) Kolbe, H. J. Prakt. Chem. 1874, 10, 89.
(25) Hale, D. K.; Hawdon, A. R.; Jones, J. I.; Packham, D. I. J. Chem. Soc.
1952, 3503.
(26) Becker, H. G. O. et al. Organikum; Deutscher Verlag der Wissenschaften:
Berlin, 1986; pp 334-336.
(27) Oddo, G.; Mameli, L. Gazz. Chim. Ital. 1901, 31, 244.
(22) Schmitt, R. J. Prakt. Chem. 1885, 31, 397.
480
•
Vol. 9, No. 4, 2005 / Organic Process Research & Development

on the Kolbe-Schmitt synthesis32 and with achieving regi-
oselective carboxylation.33
3. Aqueous Kolbe-Schmitt Synthesis Using Resorcinol in
the Microreactor Rig
In this paper, the reaction of resorcinol to 2,4-dihydroxy-
benzoic acid in aqueous solution at high temperatures and
high pressure using a microreactor rig is reported34 (compare
refs 24-26). This process was taken as a model reaction to
explore the possibility to significantly reduce the reaction
times and, respectively, to increase the space-time yields by
operation in a microreactor rig at temperatures much above
the normal level. This is seen in the context of providing
specially adapted chemical protocols for novel microreactor
chemistry as outlined above.
Figure 1. Enhancement of boiling temperatures of typical
solvents for organic reactions by rising pressure. Thus, by
having overpressure, a single liquid-phase operation can be
achieved much above the boiling point.
Figure 1).35 This is especially facile for microreactor rigs
with their small internal volumes, few numbers of seals, and
low safety precautions needed.
5. Experimental Section
5.1. Construction of the Microreactor Rig. For initial
experiments for process development, a simple first-hour
process flow sheet was applied consisting only of a capillary
reactor and a micro heat exchanger for cooling the reaction
mixture which was operated with premixed reactant solutions
(see Figure 2). In later experiments for process optimisation
with premixed solutions, an advanced microreactor rig was
installed consisting of a micro preheater, used when operating
at high flow rates. Still later, another version of the
microreactor rig was equipped with a microstructured mixer
(and alternatively a T-junction) in addition so that two
separate reactant solutions could be fed.
Concerning the heating-up of reactant solutions, both
calculations and temperature measurements revealed that
solely a certain entrance section of the capillary reactor is
sufficient at low flow rates up to 500 mL/h. At higher flow
rates, a second micro heat exchanger was used, since
otherwise a notable portion of the capillary reactor would
have to be used and most of the reaction would have been
conducted under ramping and unknown temperature condi-
tions.
The first-hour microreactor rig comprised a syringe pump
(1000D; Teledyne ISCO Inc., Lincoln, NE/USA). As capil-
lary reactor, commercial bent-steel tubing was used. As
mixing tools, the IMM split-and-recombine micromixers
CPMM-R600 and CPMM-R1200 of the caterpillar series17
and a T-junction with minute internal tubing (Ø: 2.3 mm)
were applied. A simple electrically heated tube device
(abbreviated Tube HTMD)33 was selected for heating the cold
premixed solution to reaction temperature and a cross-flow
fluidically driven micro heat exchanger (CRMH)33 to quench
the hot reacted solution down to ambient temperature.
The pressure increase in the microreactor rig to overpres-
sures up to about 75 bar was achieved by a fine-metering
The Kolbe-Schmitt synthesis was selected for the fol-
lowing reasons:
Unlike many organic syntheses neither gas nor solid is
introduced or evolved. The reaction is only temperature-
driven, i.e., premixing of the reactants in an external
conventional flask can be done which renders the processing
simpler by avoiding an integrated mixing step in the first
run. Finally, the use of water as solvent stands for safe
processing with an inexpensive and harmless solvent under
the quite extreme conditions. This is particularly relevant
for a later scale-up towards throughputs in the 100 or 1000
L/h range. The reactants resorcinol and potassium bicarbonate
are cheap and available in large quantities. Actually, purum-
grade (and not puriss -grade) qualities were used for all
experiments to be a bit closer to real-word technical
applications.
4. High p,T-Processing in Microreactor Rigs
4.1. Increasing the Temperature Processing Window
for Liquid Single-Phase Operation. Most of the laboratory
organic syntheses are performed under reflux operation at
ambient pressure in multiple-necked stirred glass flasks and
thus have an upper limit for setting the reaction temperature
given by the boiling point of the liquid. However, chemical
processing can extend to temperatures of about 80-90 °C
for many solvents (including water) higher than the boiling
point simply by applying overpressures of only 10 bar (see
(28) Wacker, A. Br. Patent 384,619, 1932; Chem. Abstr. 1932, 27, 1895.
(29) Hales, J. L.; Jones, J. I.; Lindsey, A. S. J. Chem. Soc. 1954, 3145.
(30) Hunt, S. E.; Jones, J. I.; Lindsey, A. S.; Killoh, D. C.; Turner, H. S. J.
Chem. Soc. 1958, 3152.
(31) Markovic, Z.; Engelbrecht, J. P.; Markovic, S. Z. Naturforsch., A: Phys.
Sci. 2002, 57, 812.
(32) Rahim, A. R.; Matsui, Y.; Kosugi, Y. Bull. Chem. Soc. Jpn. 2002, 75, 619.
(33) Rahim, A. R.; Matsui, Y.; Matsuyama, T.; Kosugi, Y. Bull. Chem. Soc.
Jpn. 2003, 76, 2191.
(34) Hessel, V.; Hofmann, C.; Lo¨b, P.; Lo¨hndorf, J.; Lo¨we, H. Chem. Eng. Trans.
2005, 6, 49.
(35) Lide, D. R. Handbook of Chemistry and Physics; CRC Press LLC: Boca
Raton, 1998.
Vol. 9, No. 4, 2005 / Organic Process Research & Development
•
481

Figure 2. Process flow sheets of two of three laboratory rigs used for performing process development for the aqueous Kolbe-
Schmitt synthesis using resorcinol. (Top) Initial rig for process development fed by premixed reaction solution and without
microstructured heater. (Bottom) Advanced rig for process optimization using two separate reactant solutions and having an additional
heat-exchange function for preheating.
valve which was adjusted manually by the operator. A
manual pressure adjustment was favoured over an automatic,
control-looped setting, owing to a reduced sensitivity of the
valve flow path to fouling and blocking. This allowed the
Kolbe-Schmitt synthesis to be performed under single
liquid-phase conditions even at temperatures as high as 220
°C and above (see Figure 1) instead of having two-phase
gas-liquid processing. The latter would negatively impact
conversion by decreasing residence time through the gas
consuming much of the capillary reactor and thereby pushing
the liquid segments towards higher flow velocity. Accord-
ingly, the chosen high-pressure equipment allows long
reaction times under high p,T-operation.
CPMM R600, supposed to exhibit faster mixing, and one
with a large channel of 1200 µm, CPMM R1200. The
CPMM devices were manufactured by 3-D micromilling.
mixer type
CPMM: up-down ramped,
split-and-recombine micromixer,
caterpillar- type
stainless steel SS 316 Ti
8
2
mixer material
number of SAR steps
number of
microstructured plates
channel width
channel depth
channel length
(all 8 steps)
1.2 mm
1.2 mm
19.2 mm
5.2. Microreactor Devices Used. 5.2.1. Liquid Split-and-
Recombine Micromixer CPMM-Series. This micromixer has
a ramp-like internal microstructure (see Figure 3) within one
channel which is alternately directed up and down1 (see ref
36 for the second-generation device). This induces at low
Reynolds numbers a split-and-recombine action which is a
sequential multiplication of the number of fluid lamellae,
while halving their width36 (see also ref 37). At high
Reynolds numbers, circulatory flow gives eddies which lead
to interfacial stretching. Thus, diffusion is the major mixing
mechanism at low Reynolds numbers, while convection
(followed by diffusion) is effective at high Reynolds
numbers. Two versions of the CPMM mixer were used in
the experiments. One with a small channel of 600 µm,
mixer type
R1200
ramp width
0.6 mm
1.2 mm
0.6 mm
0.6 mm
9.6 mm
ramp length
channel width
channel depth
channel length
(all 8 steps)
mixer type
R600
ramp width
0.3 mm
ramp length
1.2 mm
O-seal material
outer device dimensions
washer FPM
2 mm × 4 mm × 96 mm
total volume flow
operational temp
system pressure
liquid mixing time
up to 70 L/h
up to 200 °C
up to 30 bar/435 psig
20-100 ms
5.2.2. Electrical Preheater Tube HTMD. This electrically
operated preheater (see Figure 3) comprises an open tube in
which a microstructured body carrying parallel straight
(36) Scho¨nfeld, F.; Hessel, V.; Hofmann, C. Lab Chip 2004, 4, 65.
(37) Hessel, V.; Lo¨we, H.; Scho¨nfeld, F. Chem. Eng. Sci. 2005, 60, 2479.
482
•
Vol. 9, No. 4, 2005 / Organic Process Research & Development

Figure 3. Microstructured devices used. (Top left) Disassembled caterpillar micromixer CPMM-R1200, mainly showing the two
microstructured housing plates. (Top right) Up-down curved ramp-type microstructured channel in the CPMM mixer. (Bottom
left) Electrically stirred liquid microstructured preheater “Tube HTMD”. (Bottom right): fluidically driven liquid microstructured
cooler “CRMH”.
channels is inserted.38 It has a simple overall architecture
with minimal bends, i.e., having two-inlet feed and only
straight channels in the interior. The device was manufac-
tured by micromilling.
guidance of the reaction and cooling fluids in a complex
multiple-stack architecture.33 The plates have multiple break-
throughs which are sealed by other plates, thereby creating
microchannels. The plates are turned by 90° each, thus giving
the cross-flow architecture. The plates were manufactured
by punching and joined by diffusion bonding (the latter done
by the Heatric Company, Poole/U.K.).
heat exchanger type
HTMD: electrically heated micro heat
exchanger tube
stainless steel SS 316Ti
32
mixer material
number of microchannels
per tube
heat exchanger type
CRMH: cross-flow micro heat
channel width
channel depth
total length
device diameter
device length
inlet diameter
outlet diameters
powerelectrical
0.5 mm
1.5 mm
60 mm
12 mm
80 mm
1/4 in.
exchanger
mixer material
number of microchannels
per plate
number of microstructured
plates
stainless steel SS 304
9
52
1/8 in.
channel width
channel depth
0.7 mm
0.2 mm
12.0 mm
200 W
channel total length
(mm)
5.2.3. Capillary reactor. Two commercial capillaries
(Harry Rieck GmbH, Hilden/Germany) were used as reactor
and delay loop.
sealing
diffusion bonding
22 mm × 22 mm × 14 mm
up to 130 L/h
outer device dimensions
total volume flow
operational temperature
system pressure
heat transfer coefficient
up to 300 °C
up to 50 bar
4000 W/(m2‚K)
for water at 30 L/h
12800 m2/m3
reactor type
bend capillary
stainless steel SS 316Ti
1.3 mm
mixer material
specific heat transfer area
active inner volume
pressure drop
15 mm3
tube inner diameter
tube outer diameter
(1st capillary)
1/8 in.
1 bar at 72 L/h
watery fluid
tube inner diameter
tube outer diameter
(2nd capillary)
length (1st capillary)
length (2nd capillary)
1.76 mm
1/8 in.
5.3. Chemical Protocol. To focus on establishing the
novel means of microreactor p,T-processing, the number of
microstructured devices was reduced to those absolutely
needed in the first experimental runs. The KHCO3 concen-
tration (253 g/L) had to be set close to the solubility limit
(which is 333 g/L), and the resorcinol concentration was
similarly high. Initial experiments revealed that precipitation
4.4 m
3.9 m
5.2.4. Liquid Cross-Flow Cooler CRMH. This preheater
(see Figure 3) is fluidically driven and uses a cross-flow
(38) IMM, unpublished.
Vol. 9, No. 4, 2005 / Organic Process Research & Development
•
483

of such a concentrated solution may occur when exposed to
high temperature. Thus, the use of a micromixer was omitted
in the initial experiments, as this component is known to be
liable to clogging. Consequently, premixed solutions with
both reactants dissolved were used. This is justified, since it
was known that the reaction hardly proceeds at room
temperature.
The type of reactants and their concentrations were
oriented on a standard protocol of the German student
textbook for chemical practical courses in organic synthesis.26
• resorcinol: 73.5 g (0.782 mol)
• KHCO3: 202.0 g (2.02 mol)
• water: 669.0 g
Figure 4. Batch operation of the aqueous Kolbe-Schmitt
synthesis using resorcinol in a typical laboratory stirred vessel
under reflux conditions. Samples were taken at defined time
intervals.
The result is an 800-mL aqueous stock solution with a
0.98 mol/L concentration of resorcinol. The textbook’s
recommendation, a 4-5-fold excess of KHCO3 to resorcinol,
was not used; instead was used only a 2.5-fold excess to
avoid particle precipitation and clogging of the microreactor
rig. Temperatures (100-220 °C), overpressures (1-74 bar),
and reaction times (4-390 s) were set as given below.
For the later experiments using the caterpillar micromixer,
two solutions of the following composition were fed.
• solution 1: 147 g of resorcinol (1.564 mol) in 105 g of
water
Table 1. Impact of overpressure on yield of the aqueous
Kolbe-Schmitt synthesis using resorcinol, validating the
high-p,T processing microreactor concept and sketching the
threshold pressure value for stable single-liquid phase
operation
temp
[°C]
flow rate
[mL/h]
reaction time
[s]
overpressure
[bar]
yield
[%]
120
120
120
120
84
84
84
84
390
390
390
390
0.3
8
16
32
23
46
45
47
• solution 2: 404 g of KHCO3 (4.04 mol) in 1233 g of
water
Most of the solvent water was given to solution 2 (owing
to the low solubility of KHCO3), resulting in an asymmetric
ratio of about 10 for the two individual flow rates. While
the function of many micromixers may be reduced at such
flow ratios, the mixing efficiency of the caterpillar micro-
mixer is not altered.
were taken from a stirred 1000-mL laboratory flask at defined
intervals and analysed. In this way, it was, for example, found
that a 40%-yield at a selectivity of about 90% was obtained
after 120 min of reaction time (see Figure 4). In the literature,
a maximum yield of 50% is reported after 2 h of operation
so that in this case the protocol operation time was set quite
accurately. This also means that it is not likely that a
microreactor operation will give better results when keeping
the same temperature and reaction time settings. On the
contrary, other protocol settings have to be found for the
microreactor to yield process intensification. In addition, the
selectivity was determined using the batch data. A selectivity
of about 90% to the product was found when the reaction
was completed.
The first-hour microreactor experiments were made at a
reaction time of 390 s (equalling a total flow rate of 84 mL/
h), respectively 6.5 min, and a temperature of 120 °C, varying
the overpressure from 0.3 to 32 bar. On the basis of this
information and the fact that approximately a doubling of
the reaction time may be expected for each 10 °C temperature
rise (i.e., 16-fold at 140 °C, when 100 °C is the reference
temperature), the very first microreactor experiments were
made at a reaction time of 706 s (equalling a total flow rate
of 50 mL/h), respectively nearly 12 min, and a temperature
of 140 °C. Actually, much longer reaction times are hardly
feasible and useful, as this implies usage of still longer
reaction capillaries, respectively higher pressure drops, and
reduced space-time yields.
5.4. Materials. 2,4-Dihydroxybenzoic acid (C7H6O4,
154.12 g/mol, Fluka, g98%), resorcinol (C6H6O2, 110.11
g/mol, Aldrich, 99%), and potassium hydrogen carbonate
(KHCO3, 100.12 g/mol, Fluka, g98%) were used. The water
was deionised by double distillation. The aqueous solutions
were degassed by exposure to ultrasound and subsequently
bubbled with nitrogen to remove the oxygen contents before
use, because dissolved oxygen could lead to undesired
oxidation reactions at the high temperatures envisaged.
5.5. HPLC Product Analysis. The contents of the
reaction mixture were determined by HPLC (Shimadzu VP
series with UV-vis detector) with acetonitrile (Aldrich) as
eluent and the following chromatographic conditions: iso-
cratic flow 0.6 mL/min, (50/50) CH3CN/0.5 mol/L KH2PO4
in 1% H3PO4, 250 mm × 4.0 mm Nucleosil 120 C18, 5 µm
column (Macherey Nagel, Du¨ren/Germany purchased via
MZ-Analysentechnik, Mainz/Germany), and UV detection
at 220 and 260 nm.
6. Results and Discussion
6.1. Process Development. A time of 2 h is reported for
completion of the reaction when operating in a laboratory-
scale flask.25,26 However, this setting has to be questioned,
as most of the organic protocols do not refer to intrinsic
kinetics and thus are set (by far) too long. For this reason, a
kinetic study for the batch operation was performed to get a
more exact estimation of the actual reaction time. Samples
It was found that an overpressure of about 8 bar or more
is required for stable and efficient operation. The following
484
•
Vol. 9, No. 4, 2005 / Organic Process Research & Development

Table 2. Optimised protocol settings for the aqueous
Kolbe-Schmitt synthesis using resorcinol in the
microreactor rig: reaction temperatures being further
increased, flow rates being set higher, respectively shorter
residence times, and overpressures are increased to about 40
bara
temp
[°C]
flow rate
[mL/h]
reaction time
[s]
overpressure
[bar]
yield
[%]
120
140
140
140
140
140
140
160
160
160
160
160
180
200
200
200
200
200
200
200
220
500
84
250
65
390
130
65
32
16
11
130
65
32
16
11
65
130
65
32
16
11
8
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
40
12
47
35
28
12
4
500
Figure 5. Yields for the aqueous Kolbe-Schmitt synthesis
using resorcinol in the microreactor rig: variation of reaction
temperature at low flow rates, respectively long residence times,
and moderate overpressures.
1000
2000
3000
250
1
38
37
15
5
500
experiments being conducted at 8 bar show that the yield
increases with reaction temperature (see Figure 5). At 140
°C, the continuous-flow microreactor performance (yield:
48%) approaches that of the batch, albeit at much reduced
reaction times. However, fouling and clogging (most likely
by resorcinol- and KHCO3 precipitation) were significant at
these low flow rates. This was a result of unstable plant
operations during part of the time and also because of less
accurate and reliable analytical data concerning the product
formation. The determination of the resorcinol conversion
differed under seemingly the same conditions, and thus the
analysis was only based on determining the product yield.
In consequence, higher flow rates at still higher temperatures
and overpressure were chosen for the next experiments. For
these reasons, the plant rig was changed as well to a more
robust version, less prone to fouling; the choice of the Tube
HTMD micro heat exchanger especially constituted a major
improvement.
1000
2000
3000
500
250
500
1000
2000
3000
4000
5000
500
1
43
34
35
34
38
17
10
8
6
65
31
a The respective yields are indicated.
initially generated. At high temperatures the γ-resorcylic acid
readily decomposes to resorcinol and carbon dioxide.25
6.2. Process Optimization. 6.2.1. In-depth InVestigation
of Residence Time and Temperature Effects. In the second
part of experiments, a more systematic variation of process
parameters such as residence time and temperature was
undertaken for a large parameter range from 6 to 560 s and
100 to 220 °C (see Table 2).
For the salicylic acid synthesis, a mechanistic analysis
was made to determine how such arrangements of the
hydroxyl group may occur.33 It is expected that this proceeds
in a similar way for the â-resorcylic acid.
The yields approached and even exceeded 40% (e.g., 160
°C, 40 bar, 500 mL/h, 65 s or 200 °C, 40 bar, 2000 mL/h,
16 s; see Table 3). Thus, the performance of the batch
operation (yield: 59%) was nearly reached.26 The two
microreactor operations mentioned above, accordingly, amount
to a 110- or 450-fold reduction in reaction time as compared
to standard batch-operation (see Table 3 and Figure 5).
In some sources other than ref 26, higher yields up to
60% are reported,25 especially in the presence of carbon
dioxide gas in addition to the carbonating KHCO3. It is also
cited basically for all variants of the Kolbe-Schmitt synthesis
that both â-(2,4-di-hydroxy)- and γ-(2,6-di-hydroxy)-
resorcylic acids are formed.25 The exact share of both isomers
is mainly determined by the reaction time and temperature.
Generally, long reaction times (some 10 h) and high
temperatures (>150 °C) favour the formation of the γ-re-
sorcylic acid, the content of which then can amount up to
50%.25 It is usually assumed that the γ-resorcylic acid is
formed by follow-up rearrangement of the â-resorcylic acid
It is also known that γ-resorcylic acid decomposes to
resorcinol and carbon dioxide.25 Other mechanistic investiga-
tions revealed how such thermal decomposition occurs for
the potassium and other salts of salicylic acid.30
In view of this information, it is likely that at the high
temperatures used some of the product â-resorcylic acid
actually converted to γ-resorcylic acid and finally was
decomposed to resorcinol. As a net effect, the yield may
Vol. 9, No. 4, 2005 / Organic Process Research & Development
•
485

Table 3. Optimised protocol settings for the aqueous Kolbe-Schmitt synthesis using resorcinol in the microreactor rig:
reaction temperatures being further increased, flow rates being set higher, respectively shorter residence
normalized
assumed processed
flow rates [mL/h]
concentration
[g/L]
productivity
[g/h]a
space-time yield
space-time yield
type of processing
[kg/h‚m3]
[-]
microreactor processing:
single capillary
2000
88
111
12,330
440
(9 mL inner volume)
batch processing (1 L flask)
microreactor processing:
5 parallel capillaries
(9 mL inner volume)
500 () 1000 mL/2 h)
2000
88
88
28
555
28
12,330
1
440
a Assuming 45% yield.
Table 4. Protocol settings and yields for the aqueous Kolbe-Schmitt synthesis using resorcinol in the microreactor rig using
three mixing tools, the caterpillar micromixers CPMM 600 and CPMM 1200 and a commercial T-junction
temp
[°C]
total flow rate
[mL/h]
individual flow rates
KHCO3/resorcinol [mL/h]
reaction time
[s]
overpressure
[bar]
yield
[%]
mixing tool
CPMM-R600
CPMM-R600
CPMM-R600
CPMM-R600
CPMM-R600
CPMM-R1200
CPMM-R1200
CPMM-R1200
T-junction
160
160
160
160
160
160
160
160
160
160
100
200
500
1000
2000
500
1000
2000
500
86/14
172/28
430/70
861/139
1721/279
430/70
861/139
1721/279
430/70
325
162
65
32
16
65
32
16
65
40
40
40
40
40
40
40
40
40
40
48
38
44
37
28
46
41
27
43
43
T-junction
1000
861/139
32
have been decreased and this is why not much more than
about 45% yield was reached at maximum; all the γ-resor-
cylic acid contents which were generated may have been
vanished in situ. A simple control experiment would be the
guidance of commercially purchased â-resorcylic acid and
γ-resorcylic acid under the high p,T-conditions through the
microreactor rig and comparison of initial and final concen-
trations.
6.2.2. ProductiVity and Space-Time Yield. From the yield
and other data given in the last chapter, an increase in the
space-time yield by microreactor operation can be confirmed
(see Table 4). A 440-fold increase as compared to the batch
performance (1-L-flask; based on data from ref 8) can be
stated, assuming a flow rate of 2000 mL/h for the microre-
actor and a yield of 45%. This is, to a considerable extent,
owing to the small inner volume of the microreactor capillary
(9 mL), whereas the outer footprint of the microreactor plant
is certainly not much below that of the stirred 1-L batch
reactor. Thus, the productivity in terms of material produced
per time is probably the more relevant figure of merit. Such
microreactor productivity, using the current settings, amounts
to 111 g/h, which is 4 times that of the batch. Table 3 also
gives an outline for productivity to be reached with five
capillary reactors operated in parallel, which is feasible
according to in-house experience. Then, a productivity of
about 0.55 kg/h is achieved, respectively 13.2 kg per day
and 4.4 t per year (assuming 24-h daily operation and 8000
h per year, excluding maintenance).
Figure 6. Impact of setting the residence time via modifying
the total flow rate on the yield of the aqueous Kolbe-Schmitt
synthesis using resorcinol in the microreactor rig.
the residence time is given for three selected temperatures
of 140, 160, and 200 °C.
The highest productivity is given at a temperature of 200
°C and a flow rate of 2000 mL/h with a yield of 30% where
the interplay between throughput and conversion is maximal
6.2.3. Variation of Flow Rate and Residence Time. In
Figure 6, the dependency of the yield on the flow rate and
486
•
Vol. 9, No. 4, 2005 / Organic Process Research & Development

Figure 8. Yields of the aqueous Kolbe-Schmitt synthesis using
resorcinol, when different mixing tools are incorporated into
the microreactor rig, placed in front of the capillary reactor.
CPMM R600: caterpillar micromixer with 600-µm inner
channel; CPMM R1200: caterpillar micromixer with 1200-µm
inner channel; commercial T-junction with 2.3-mm inner
channel.
Figure 7. Impact of setting the residence time via modifying
the total flow rate on the yield of the aqueous Kolbe-Schmitt
synthesis using resorcinol in the microreactor rig.
synthesis. Large asymmetric flow ratios were used because
of very different solubilities of the reactants (see section 4.3.).
Figure 8 shows the yield as a function of the total volume
flow at a temperature of 160 °C, when the different mixing
tools are used (see also Table 4).
(see Figure 5). At total flow rates above 2000 mL/h, the
reaction time provided is not sufficient anymore, resulting
in incomplete conversion. Below 2000 mL/h at a too-long
residence time, the yield decrease may refer to selectivity
losses, i.e., by increasing the content of side and consecutive
reactions.
6.2.4. Variation of Temperature. In Figure 7, the depen-
dency of the yield on the reaction temperature is plotted for
various flow rates against residence times. It is evident that
the yield increases largely with temperature up to about 180
°C for all data sets.
For all three devices, a decrease in yield can be observed
with increasing total volume flow, due to the reduction of
residence time. No major difference between the two
microstructured mixers and the T-junction is given, at least
in the flow range investigated here. This is not surprising as
the T-junction is probably operated in the “intertwined”
regime where the streamlines of the two solutions to be
mixed are effectively entangled.39-41 Probably the perfor-
mance of the T-junction would decline for operation at low
Reynolds numbers (here, at still lower total volume flow),
because then a “stratified” flow is exhibited, similar to a bi-
laminated fluid system with large diffusion distances and
respectively inefficient mixing.
Bearing in mind that mixing was performed here using
largely asymmetric flow ratios of about 10, the results also
show that micromixers can be effectively used for performing
the aqueous Kolbe-Schmitt synthesis even under these
constraints. No loss in efficiency is observed as compared
to the option with premixed solutions.
6.3.2. PreparatiVe Product Isolation, Purification, and
Identification. The crude product (from one experiment with
the CPMM-R600 mixer at 1000 mL/h, 160 °C, 40 bar) was
isolated by dropping the reaction solution into concentrated
HCl solution.26 The precipitated crystals were filtrated. Then,
the crude product was heated with a small volume of water
in a stirred flask with reflux condenser until boiling was
achieved. The suspension was fed slowly with water until
complete dissolution of the solid is achieved. Heating and
stirring were stopped immediately, and the clear solution was
slowly cooled without any shaking. Precipitates were formed
and filtrated again. The purified product is composed of white
crystals (215 °C, decomposition) which turn slightly pink
after some days due to oxidation of the electron-rich aromatic
core. A yield of about 22% of the purified product (2.59 g)
For one data set (500 mL/h) operation at still higher
temperatures was performed as well. The performance
decreased for the reasons given above. The possible relation-
ship for the other four curves (1000 to 4000 mL/h) is less
clear, when considering the few data points gathered. These
curves seem to have a different dependency, as they have a
much larger gradient of the yield increase; however, caution
must be taken concerning too far-fetched conclusions,
because of the small number of data points and the problems
encountered with stable operations without any precipitation.
6.3. Processing with Online Mixing by Micromixers.
6.3.1. In-depth InVestigation of Residence Time and Tem-
perature Effects. The use of premixed solutions poses limits.
Most important, offline mixing creates safety constraints the
larger these volumes become. Therefore, such operation is
certainly not acceptable at a production level. Second, one
cannot completely exclude pre-reactions such as with oxygen.
Moreover, a simple screening of the concentrations of the
reactants is not possibl; rather, manifold stock solutions
would have to be prepared. Finally, a transfer of the high-
p,T processing developed here towards other organic reac-
tions, which have mixing sensitivity, would not be easily
possible.
For all these reasons, it was tested if the implementation
of a micromixer into the rig would give as good or even
better performance. Because it was recently found that
T-junctions with minute diameter may have a mixing
performance similar to that of micromixers,39-41 this option
was tested as well. Actually, two micromixers of the
caterpillar series were tested with differing internals. It was
tested how the choice of a mixer with a larger inner diameter,
which is the means to match an increase in productivity
demands, may affect the performance of the Kolbe-Schmitt
(39) Wong, S. H.; Ward, M. C. L.; Wharton, C. W. Sens. Actuators, B 2004,
100, 359.
(40) Wong, S. H.; Bryant, P.; Ward, M.; Wharton, C. Sens. Actuators, B 2003,
95, 414.
(41) Engler, M.; Kockmann, N.; Kiefer, T.; Woias, P. Chem. Eng. J. 2004, 101,
315.
Vol. 9, No. 4, 2005 / Organic Process Research & Development
•
487

Figure 9. 1H NMR spectra (solvent d6-DMSO) of the purified product synthesized by the aqueous Kolbe-Schmitt synthesis in the
microreactor rig. No side products are detectable.
was obtained, which is considerably lower than the analytical
HPLC yield of 40% and the literature data of 50%.
Obviously, some losses of the product â-resorcylic acid
occurred during precipitation. Further optimization of the
purification step was not an objective within the works
reported in this paper. Probably the use of additional
extraction steps with diethyl ether will help here.25
FTIR (Nicolet Magna IR 750, KBr, cm-1): ν(ring
C-H) 1532, ν(CdO) 1885, ν(CO-OH) 3286, ν(dC-OH)
3640.
GC-MS (Agilent 6890 GC/5973N MSD, EI, m/z): 136
(C7H4O3); 110 (C6H6O2); 81 (C5H5O); not found: 154
(C7H6O4, product peak), since all product was decarboxylated
under the MS conditions.
The chemical nature of the product was confirmed in
detail by 1H NMR, 13C NMR, FTIR, GC-MS, and elemental
analysis. The NMR spectra show no traces of other organic
molecules, in particular demonstrating that the reactant
resorcinol was effectively separated (see Figures 9 and 10).
The elemental analysis with regard to C and H, however,
differs notably from the calculated data. This is explained
by the remainder of KHCO3 in the purified product, which
coprecipitates with the â-resorcylic acid. Since the product
formation was undoubtedly identified by the other methods,
no further attempts to remove the KHCO3 contents were
undertaken within the works reported here.
1H NMR (Bruker AMX 400, 400 MHz, d6-DMSO,
δ/ppm): 6.2 (s -C(OH)-CHdC(OH)- 1H); 6.4 (d -CHd
C(OH)-CHdC(OH)- 1H); 7.6 (d -CHdC(COOH)- 1H);
10.4 (s -CHdC(COOH)- 1H); 10.4 (s dC(COOH)-
C(OH)-,1H); (-C(OH)-CHdC(OH)-, hidden by broad
peak at 3.4 of H2O traces within the solvent DMSO).
13C NMR (Bruker AMX 400, 100.6 MHz, d6-DMSO,
δ/ppm): 102.3 (-C(OH)-CHdC(OH)-), 104.3 (-CHd
C(COOH)-), 108.0 (dCH-CHdC(COOH)-), 131.9 (-C-
HdC(COOH)-), 163.4 (-CHdCH-CHdC(COOH)-),
164.1 (dC(COOH)-C(OH)-), 172.0 (dC(COOH)-).
7. Conclusions and Outlook
The advantages of the new microreactor synthesis for the
aqueous Kolbe-Schmitt synthesis compared to a 1-L labora-
tory flask synthesis7,8 are as follows.
• reduction of reaction time by orders of magnitude (few
tens of seconds instead of minutes)
• increase of space-time yield by orders of magnitude
• increase of throughput by a factor of 2 (with option to
one magnitude by numbering-up)
• simple and flexibly upgradeable rig, for laboratory and
pilot throughputs
The yield approached data reported in the literature.8 The
process benefits are related to the simple high-p,T operation
and the defined setting of short residence times using steep
temperature ramps for preheating and cooling. Advantages
in terms of mixing and mass transfer are not evident.
The disadvantages of the new microreactor technique are
the following:
• partly unstable plant operation due to pronounced
sensitivity to fouling
• unreliable resorcinol analysis due to resorcinol deposits
and decomposition reactions in the plant
488
•
Vol. 9, No. 4, 2005 / Organic Process Research & Development

Figure 10. 13C NMR spectra (solvent d6-DMSO) of the purified product synthesized by the aqueous Kolbe-Schmitt synthesis in
the microreactor rig.
• capital and energy expenditure for high temperature and
pressure operation
For optimisation of the microreactor-based aqueous
Kolbe-Schmitt synthesis, credit should be given to the
following topics, which actually will be investigated in the
follow-up R&D.
• batch to continuous flow
• variable throughput, production on-demand
• reduction of plant footprint size with the option of
mobile transfer for distributed, delocalised production
• modular system with adaptability towards other custom-
ized synthesis
• finding of additives to the solvent (e.g., cosolvent) for
better dissolution of the reactants, in particular concerning
the KHCO3
• change to a more soluble carbonating species than
KHCO3
• feed of CO2 as another carbonating agent
• automated pressure build-up and control
• construction of scale-out solutions with several capil-
laries in parallel
The development of a five-parallel-microreactor capillary
plant with one large-throughput microstructured CPMM
mixer is actually under way at IMM.
Once these process optimisations have been realised, it
may be argued about benefits in terms of whole-plant
operation and finally with regard to economic cost calcula-
tionsif not of interest for the Kolbe-Schmitt synthesis then
certainly for another, similar, high-p,T microreactor process.
Generically speaking, the following process innovation links
may serve as business arguments for the use of the new
technique.
• possible reduction in capital costs for equipment since
a few inexpensive small-sized parts manage high throughputs
The following business drawbacks concerning microre-
actor plant operation are seen.
• possibly more dedicated process control loops, in
particular to monitor stable flows and fouling
• possibly higher energy costs, in particular for building
up high overpressures
• possibly reduction of the concentration and thus a larger
share of solvent (which is, however, water only)
Acknowledgment
We thank in particular Dorothee Reinhard for performing
the HPLC analyses. Acknowledgments are also given to Dr.
Heinz Kolshorn and Mr. Diendorf (Department of Organic
1
Chemistry, University of Mainz) for the H/13C NMR and
elemental analysis measurements, respectively.
Received for review March 18, 2005.
OP050045Q
Vol. 9, No. 4, 2005 / Organic Process Research & Development
•
489