Direct fluorination of toluene using elemental fluorine in gas/liquid microreactors

Article abstract of DOI:10.1016/S0022-1139(00)00300-6

Direct fluorination of toluene, pure or dissolved in either acetonitrile or methanol, using elemental fluorine was investigated in gas/liquid microreactors, namely a falling film microreactor and a micro bubble column. The experiments included measurements at high substrate concentrations and at high fluorine contents diluted in a nitrogen carrier gas, e.g. up to 50vol.% fluorine. Results obtained were compared to the performance of a laboratory bubble column which served as a technological benchmark. Due to the formation of liquid layers of only a few tens of micrometers thickness, the microreactors provide very large interfacial areas, e.g. up to 40,000m²/m³. These values exceed by far those of the laboratory bubble column as well as all other devices applied in practice. The potential for enhancing mass and heat transfer was verified by several experiments resulting in an increase in conversion and selectivity for the microreactors compared to the laboratory benchmark. For the falling film microreactor, yields of up to 28% of monofluorinated ortho and para products for a degree of toluene conversion of 76% were obtained. These values are of the same order as described for the industrially applied Schiemann process. Space-time yields of the microreactors, when referred to the reaction channel volume, were orders of magnitude higher than those of the laboratory bubble column. Taking into account the construction material needed, the corresponding figures of merit, for an idealized geometry as well as the existing total reactor geometry, still indicate technological and economic benefits. A variation of operating conditions for the direct fluorination revealed that conversion can be increased in the microreactors by using higher fluorine-to-toluene ratios and reaction temperatures. The choice of solvent is also essential, with acetonitrile yielding much better results than methanol.

Full text of DOI:10.1016/S0022-1139(00)00300-6

Journal of Fluorine Chemistry 105 (2000) 117±128
Direct ¯uorination of toluene using elemental ¯uorine
in gas/liquid microreactors
a,*
K. Jahnisch , M. Baerns , V. Hessel , W. Ehrfeld , V. Haverkamp ,
a
b,*
b
b
È
b
H. Lowe , Ch. Wille , A. Guber
b
c
È
Institut fur Angewandte Chemie Berlin-Adlershof e.V., Richard-Willstatter-Str. 12, D-12489 Berlin, Germany
a
È
b
È
È
Institut fur Mikrotechnik Mainz GmbH, Carl-Zeiss-Str. 18-20, D-55129 Mainz, Germany
c
È
Forschungszentrum Karlsruhe GmbH, Institut fur Mikrostrukturtechnik, Postfach 3640, D-76021 Karlsruhe, Germany
Received 7 February 2000; accepted 2 May 2000
Dedicated to late Professor Schiemann on the occasion of the 100th anniversary of his birth
Abstract
Direct ¯uorination of toluene, pure or dissolved in either acetonitrile or methanol, using elemental ¯uorine was investigated in gas/liquid
microreactors, namely a falling ®lm microreactor and a micro bubble column. The experiments included measurements at high substrate
concentrations and at high ¯uorine contents diluted in a nitrogen carrier gas, e.g. up to 50 vol.% ¯uorine. Results obtained were compared
to the performance of a laboratory bubble column which served as a technological benchmark.
Due to the formation of liquid layers of only a few tens of micrometers thickness, the microreactors provide very large interfacial areas,
e.g. up to 40,000 m2/m3. These values exceed by far those of the laboratory bubble column as well as all other devices applied in practice.
The potential for enhancing mass and heat transfer was veri®ed by several experiments resulting in an increase in conversion and
selectivity for the microreactors compared to the laboratory benchmark. For the falling ®lm microreactor, yields of up to 28% of
mono¯uorinated ortho and para products for a degree of toluene conversion of 76% were obtained. These values are of the same order as
described for the industrially applied Schiemann process. Space-time yields of the microreactors, when referred to the reaction channel
volume, were orders of magnitude higher than those of the laboratory bubble column. Taking into account the construction material needed,
the corresponding ®gures of merit, for an idealized geometry as well as the existing total reactor geometry, still indicate technological and
economic bene®ts.
A variation of operating conditions for the direct ¯uorination revealed that conversion can be increased in the microreactors by using
higher ¯uorine-to-toluene ratios and reaction temperatures. The choice of solvent is also essential, with acetonitrile yielding much better
results than methanol. # 2000 Elsevier Science S.A. All rights reserved.
Keywords: Fluorination; Toluene; Elemental ¯uorine; Falling ®lm microreactor; Micro bubble column
1. Introduction
by a gas phase process as described by Schiemann et al. in
1968 [2]. However, selectivities turned out to be very low.
Other authors investigated direct ¯uorination of aromatic
compounds in the liquid phase. In the very early attempts
using conventional reactors, direct ¯uorination experiments
were characterized by the formation of unde®ned products
and often led to explosions [3] due to too large heat release.
In the last 30 years, substantial efforts have been focused on
the development of more advanced processing routes, e.g.
improved heat transfer characteristics, allowing the desired
electrophilic substitution path to proceed. By operation at
extremely low temperatures, e.g. 708C, and reactant con-
centrations, e.g. down to 0.01 mol/l, the proof of feasibility
for performing direct ¯uorination was given for aromatic
substrates like benzene, toluene and phenol [4±7].
The direct ¯uorination of aromatics with pure ¯uorine on
a technical scale is a problem still unsolved to date. For the
synthesis of the vast majority of ¯uorinated compounds
multi-step processes such as the Schiemann reaction are
employed. In this process, aromatic amines are converted
into diazonium tetra¯uoroborates, which thermally decom-
pose with formation of mono¯uorinated aromatic com-
pounds in reasonable yields [1].
On the laboratory scale, a number of experiments using
elemental ¯uorine as a ¯uorination agent have been carried
out. For instance, the ¯uorination of toluene was performed
* Corresponding authors. Fax: ‡49-61-31-990-201.
0022-1139/00/$ ± see front matter # 2000 Elsevier Science S.A. All rights reserved.
PII: S 0 0 2 2 - 1 1 3 9 ( 0 0 ) 0 0 3 0 0 - 6

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
118
Fig. 1. Schematic diagram of a falling film microreactor, a micro bubble column and a laboratory bubble column. (a) Vertically oriented falling film
microreactor, a thin liquid film is fed through orifices into the reaction channels, generating a relatively large surface for contact to the gas. (b) Micro bubble
column the liquid and gaseous reactants are contacted through a static micromixer, and, subsequently, fed into the reaction channels. (c) In the laboratory
bubble column the gas is dispersed through a porous frit into the liquid.
Even greater control over reaction temperature and, in
addition, reactant mixing should be achievable by using
microfabricated reaction devices, so-called microreactors,
since high gradients for heat and mass transfer can be
utilized, due to small diffusion distances within the micro-
channels [8]. These transport enhancements, in general,
allow improvement in the performance of reactions which
are not carried out under a kinetically controlled regime, e.g.
fast reactions coupled to fast side reactions. If the reaction is
run in a multiphase system, the potential bene®t of using
microreactors may even be ampli®ed. This scenario ideally
holds for gas phase reactions on solid catalysts and the
resulting improvements have been demonstrated recently by
many authors and for a large variety of applications [9,10].
The criteria sketched above obviously are also met by gas/
liquid reactions. Two approaches paved the way for further
investigations in this ®eld. In the ®rst reference, a selection
of possible microreactor designs for ef®cient gas/liquid
contacting was presented and their potential applications
were outlined [11]. The second paper aimed at giving a
sound theoretical basis for this ®eld by analyzing various
cases of diffusion hindrance [12]. This was completed by a
®rst proof of feasibility using existing contacting compo-
nents, originally designed for liquid/liquid extraction. Based
on this knowledge, the ®rst integrated gas/liquid microreac-
tion systems were built, termed a falling ®lm microreactor
(FFMR) and a micro bubble column (MBC) [11,13] (see
Fig. 1a and b) made of stainless steel. In the meantime
Chambers et al. [14] have successfully tested a simple
microreactor for the direct ¯uorination of 1,3-dicarbonyl
compounds, but no yield or selectivity data have been
reported.
The FFMR, developed for this study, generates a thin
falling ®lm of several 10 mm thickness ¯owing by means of
gravity forces. This is performed on a platelet comprising a
large number of microchannels of 100 mmÂ300 mm cross-
section. By changing volume ¯ow, control can be exerted
over ®lm thickness. The micro bubble column consists of a
mixing and a reaction unit. Various mixers containing
interdigital channel arrays of dimensions from a few micro-
meters to a few tens of micrometers were tested. For the
experiments described in this article, only one device was
used equipped with 20 mm deep gas and liquid feeding
channels 7 and 20 mm wide, respectively. For the reaction

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
119
unit comprising an array of parallel microchannels, two
types were utilized, namely platelets of 50 mmÂ50 mm
and 300 mmÂ100 mm channel cross-section, in the follow-
ing referred to as `narrow' and `wide' channels. More details
on design criteria and microreactor assembly are given in
reference [13].
It is the aim of the following discussion to give a detailed
analysis of the reaction performance of the microreactors
with regard to direct ¯uorination of toluene and to compare
the respective results to technical benchmarks. In addition, it
is aimed to gather ®rst hints for a performance optimization
of the microreactors by means of variation of operating
conditions. In this context, considerable efforts were made in
order to get precise analytical data on conversion, yield, and
selectivity.
Conversion, yield and selectivity were not referred to the
non-stoichiometric component, i.e. ¯uorine, although in
most experiments ¯uorine-to-toluene ratios below one were
used. This was caused by a lack of experimental instrumen-
tation for a quantitative determination of excess ¯uorine
remaining in the gas phase after reaction. Therefore, the
corresponding de®nitions had to be based on toluene:
nToluene;0 nToluene;R
X ˆ
Y ˆ
S ˆ
 100 ꢀ%†;
nToluene;0
nmono-F-toluene
nToluene;0
 100 ꢀ%†;
nmono-F-toluene
nToluene;0 nToluene;R
 100 ꢀ%†
where X the is conversion (%); S is the selectivity (%); Y is
the yield (%); nToluene, 0 is the amount of toluene before
reaction (mol); nToluene, R is the amount of toluene after
reaction (mol); and nmono-F-toluene is the amount of mono-
fluorinated products (ortho, meta, para) (mol).
2. Experimental
The majority of data presented were based on single
experiments. For control of accuracy, two consecutive mea-
surements were performed for selected experiments. These
measurements con®rmed a reproducibility of the experi-
ments within a range of about 5%.
During the experimental investigations, considerable
efforts had to be made to establish suitable analytical
methods for gathering reliable data. For series A, it turned
out that by using high performance liquid chromatography
(HPLC) and gas chromatography (GC) most of the reaction
products could be identi®ed. Moreover, quantitative data on
toluene consumption and generation of mono¯uorinated
products were obtained. In case of series B, fourier trans-
form infrared spectroscopy (FTIR) analysis based on a
Fluorinations were performed at two sites, namely the
Institut fuÈr Angewandte Chemie at Berlin-Adlershof (ACA)
and the Institut fuÈr Mikrostrukturtechnik in Karlsruhe. Using
the equipment at the former site, only ¯uorinations with
¯uorine contents up to 10 vol.% could be carried out, whilst
at Karlsruhe much higher contents could be utilized. Due to
this difference in operating parameters, the corresponding
measurements have been split in two groups, named series A
and B. A complete list of all experimental parameters in both
experimental series is given in Tables 1±4.
In all experiments, the liquid volume ¯ow in the micro-
reactors was controlled using a HPLC pump (L-6200A).
Temperature was monitored by resistance thermometers and
the gas ¯ow was set by mass ¯ow controllers.
Table 1
Series A: FFMR
Experimental
no.
Molar ratio mol
F2/mol C7H8
(%)
Liquid volume flow
(1.1 mmol C7H8 in
1 ml solvent) (ml/h)
Temperature
(8C)
Conversion
X (%)
Yield
Y (%)
Selectivity
S (%)
Acetonitrile
3D2
3D3
40
54
92.5
40
40
20
20
40
11.6
11.6
11.6
11.6
11.6
19.6
19.6
19.6
16
16
16
21
42
15
15
15
30
38
50
23
15
15
14
29
11
14
20
12
5
37
37
40
52
33
53
57
45
3D4
3G1
3G2
3C1
8
3C2
8
3C3
13
Methanol
4D1
20
19.6
19.6
19.6
19.6
11.1
11.1
11.1
17
17
17
17
17
17
17
15
36
12
24
42
18
34
3
5
20
14
58
42
9
4D3
30
4D2
4D4
40
54
7
10
4
4E2
40
4E3
54
9
50
41
4E4
92.5
14

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
120
Table 2
Series A: MBCa
Experimental
no.
Molar ratio mol
F2/mol C7H8
(%)
Liquid volume flow
(1.1 mmol C7H8 in
1 ml CH3CN) (ml/h)
Gas volume flow
(10% F2 in N2)
(ml/min)
Conversion
X (%)
Yield
Y (%)
Selectivity
S (%)
Version I
3L4, 304
3M4, 3P4
3N4, 3Q4
3R4
20
40
54
83
13
13
13
13
12.1
24.2
32.7
50.0
4
12
26
28
3
6
75
50
42
21
11
6
Version II
3R2/3T3/3V3
3R5/3T4/3V5
3R7/3T6/3V7
3U3/3V9
20
40
54
83
13
13
13
13
12.1
24.2
32.7
50.0
9
18
25
41
2
5
22
28
28
27
7
11
a LM: acetonitrile; temperature: 158C.
Table 3
Series A: LBCa
Experimental
no.
Molar ratio mol
F2/mol C7H8
(%)
Liquid batch
Gas volume flow
(10% F2 in N2)
(ml/min)
Conversion
Yield
Selectivity
(1.1 mmol C7H8 in
1 ml CH3CN or
CH3OH) (ml)
X (%)
Y (%)
S (%)
Acetonitrile
3E1
40
54
20
20
20
20
20
20
20
20
50
50
20
20
20
20
20
20
7
8
2
3
3
4
6
4
7
8
29
38
18
17
50
30
28
22
3E2
3H2
3H3
40
54
17
23
6
3J2/3K2
3J3/3K3
3J4/3K4
3J5/3K5
40
54
13
25
34
80
100
Methanol
4K1
4K2
40
54
20
20
50
50
6
8
1
1
17
13
a Temperature: 178C.
Table 4
Series B: FFMR
Molar ratio mol
F2/mol C7H8
(%)
Molar ratio mol
C7H8/mol CH3C
(%)
Liquid volume
flow (ml/h)
Temperature
Conversion
Yield
Selectivity
(8C)
X (%)
Y (%)
S (%)
40
100
200
100
10
10
19.6
19.6
19.6
19.6
16
16
16
16
37
62
76
7
15
19
28
3
40
31
37
43
10
100
attenuated total re¯ection (ATR) technique was applied to
quantify the product composition. This was used for mea-
surements performed at ¯uorine contents higher than
10 vol.% and substrate concentrations exceeding 1.1 mol/l.
For GC analysis, a HP 5890 instrument (Series II plus,
Hewlett Packard, Palo Alto, USA) was employed, using a
separation capillary coated with Stabilwax DA of 0.25 mm
®lm thickness. HPLC analysis was made using an instru-
ment equipped with a UV-detector L-4000A (Merck/Hita-
chi, Darmstadt, Germany). FTIR analysis was carried out
with a spectrometer IFS 88 (Bruker, Karlsruhe) including a
ATR-circular cell with a ZnSe crystal.
3. Discussion
3.1. Hydrodynamics: falling film and segmented flow
For a precise interpretation of data for direct ¯uorination,
some information with respect to the hydrodynamics of the

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
121
Table 5
Reactor type
FFMR
27,000
MBC (I)
9800
MBC (II)
14,800
Bubble columns [15,16]
50±600
Impinging jets [15,16]
90±2050
Specific interfacial area (m2/m3 reactor)
microreactors was already obtained prior to this publication
[13] and has been completed during the present study.
Speci®c interfacial areas were calculated using assumptions
from ®lm theory [15] and from measurements performed in
millimeter wide capillaries [16]. Experimental data were
gathered by means of an optical analysis technique [13] and
a chemical method, sul®te oxidation [17]. As a technical
benchmark, published data from conventional gas/liquid
contacting reactors, e.g. bubble or spray columns, were used
[18,19]. Table 5 lists data of interfacial areas measured for
the microreactors investigated and compared to the technical
benchmarks.
It was shown in previous experiments that ®lms down to
25 mm thickness can be generated in the falling ®lm micro-
reactor, using isopropanol as liquid. As shown in Table 5, the
corresponding speci®c interfacial area amounts to
40,000 m2/m3 which is far larger than for conventional
gas/liquid contacting equipment [18,19]. In the experiments
described here, acetonitrile ®lms of about 30±37 mm were
utilized corresponding to a speci®c interfacial area of about
33,000 and 27,000 m2/m3, respectively.
Investigations concerning the micro bubble column
focused on the determination of the ¯ow pattern. From
measurements in capillaries, it was likely that a segmented
¯ow pattern (referred to as Taylor ¯ow or bubble train ¯ow)
is established, i.e. bubbles of a cross-section close to that of
the channels separated by liquid plugs and surrounded by
thin liquid ®lms. First measurements using the micro bubble
column with wide channels (300 mmÂ100 mm) show that
this ¯ow pattern is found at low gas super®cial velocities
(see Fig. 2).
Increasing the gas super®cial velocity, the ¯ow pattern
changes to a gaseous core surrounded by a thin liquid ®lm
referred to as so-called annular ¯ow. Both regimes are sepa-
rated by a relatively large transition region comprising zones
of both segmented and annular ¯ow in different channels.
4. Fluorination of toluene
One series of experiments (A) was aimed at proving the
feasibility of the ¯uorination of toluene (1.1 mol toluene/1 l
solvent) using elemental ¯uorine (10% F2 in N2) in micro-
reactors. In addition, the impact of the variation of selected
process conditions on target values was investigated. In this
context, selectivity and conversion were monitored as a
function of ¯uorine-to-toluene ratio, volume ¯ow rate,
temperature, and solvent. These data were utilized to gather
selectivity-conversion diagrams and space-time yields for
comparison of reactor performance and economics. In addi-
tion to the variation of process conditions, geometric para-
meters, such as the width and depth of the microchannels of
the micro bubble column, were changed. In all these experi-
ments, data from a laboratory bubble column served as a
benchmark for standard laboratory equipment.
In another series of experiments (B), ¯uorination with
signi®cantly increased toluene and ¯uorine concentrations
was carried out. The corresponding measurements were only
performed in the falling ®lm microreactor. Tables 1±4 lists
all data from both series of experiments.
The results of toluene ¯uorination in microreactors were
compared to ¯uorination data achieved by using a macro-
Fig. 2. Drawing of a measured segmented flow pattern in the micro bubble column (MBC I) at a gas velocity of 10 ml/min. The contrast of the real
photographic image is too low for representation within this article.

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
122
Table 6
scopic gas/liquid contactor, i.e. a laboratory bubble column.
This column consists of a U-tube with a ®ne porous quartz
glass frit, and a reaction volume of 25 ml cooled by a
thermostatic bath (see Fig. 1c). Similar devices have been
used in the vast majority of relevant scienti®c investigations
described in this ®eld [4±7]. Much more elaborate conven-
tional gas/liquid devices exist which are used both on an
industrial and laboratory scale, however, the fact is that there
is no example of using such a device for direct ¯uorination.
Most likely, this is strongly prohibited for reasons of process
safety.
1
Wave number (cm )
Compound
Acetonitrile
1038, 918
1378, 736
Toluene
o-Fluorinated toluene
p-Fluorinated toluene
1232, 1186, 838, 703
1220, 1156, 813
in accordance with an electrophilic substitution mechanism
[20,21]. By increasing the ¯uorine content in the gaseous
phase from 10 to 50 vol.%, para products are predominantly
formed. By FTIR analysis, it could be shown that the
corresponding ratio of ortho to para products amounts to
approximately 1:4. The wave numbers of the detected
compounds are given for the FTIR in Table 6.
On the basis of this information, maximum yields of about
7% of the ortho isomer and 21% of the para isomer were
achieved, when applying the optimum set of process para-
meters for the falling ®lm microreactor (see series B,
Table 4). These values have to be compared to yields for
various mono¯uorinated products, including ortho- and
para-¯uoro toluene, achieved using the Schiemann reaction,
the technologically applied process [22].
Whilst the direct ¯uorination results in the predominant
formation of two isomers and non-negligible amounts of
impurities of by-products, the large-scale route via diazoti-
zation of aromatic amines yields only one isomer that alone
has to be separated from the crude product. For the latter
process, yields of this one isomer typically range from 20 to
27% which equals the sum of the yields, referring to the
three isomers, achieved in the microreactor. Thus, the
expenditure for product separation is, at present, higher
for the microreactors which, on the other hand, seem to
4.1. Product composition and substitution pattern
According to the reaction mechanisms expected, being
either radical or electrophilic [4,5], four isomers may con-
tribute, in principle, to the product composition. Radical
substitution may result in side chain or ring ¯uorination,
whereas both mechanisms lead to ortho, meta and para
isomers, usually of de®ned distribution, the so-called sub-
stitution pattern. Moreover, di- or poly¯uorinated side chain
or ring products could potentially be synthesized.
In Fig. 3, a typical GC plot corresponding to a product
mixture of the falling ®lm microreactor obtained at ¯uorine
contents of 10 vol.% is shown. Apart from the three mono-
¯uorinated isomers, peaks corresponding to a di¯uorinated
product, 2,4-di¯uoro toluene, by-products, toluene, and
acetonitrile were found. As to be expected from results
performed previously [3±7], a number of by-products were
detected, most likely generated by radical addition and
polymerisation mechanisms [4,5].
A quantitative analysis of the three well-resolved mono-
¯uorinated isomers yields a substitution pattern of 5:1:3,
with respect to the ortho, meta and para product. This ratio is
Fig. 3. A typical gas chromatography plot of a product mixture, obtained by direct fluorination of toluene in the falling film microreactor.

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
123
bene®t from simpler processing and less waste generation
compared to the multi-step technical process.
4.2. Conversion and selectivity
For various operating conditions, conversions, yields and
selectivities were determined for the micro bubble column
using two types of reaction platelets (`narrow/wide'), the
falling ®lm microreactor, and the laboratory bubble column.
In more than 20 experiments performed in these micro-
reactors, the feasibility of toluene ¯uorination could be
demonstrated by high yields of mono¯uorinated products.
Moreover, the yields of the falling ®lm microreactor exceed
for a wide range of operating conditions, with the exception
of one experiment, those of the laboratory bubble column.
The maximum yields of mono¯uorinated products deter-
mined in the two series, amounting to 20 and 28%, are more
than two to three times larger than the benchmark. The
yields obtained using the two types of platelets of the micro
bubble column do exceed this standard, but to a minor
extent.
Selectivities achieved by performing the ¯uorination in
the microreactors range from approximately 30 to 50%, i.e.
the content of by-products in the product mixture is equal
to or double that of the target products. Although such a
ratio would be unacceptable for the majority of organic
syntheses, product selectivities measured are comparable to
the best values reported for toluene ¯uorination in the
literature [4±7].
Referring to these investigations, differences of operating
conditions with respect to the present study must be men-
tioned. Only in case of using the microreactors, was proces-
sing performed at high toluene concentrations and high
molar ratios of toluene to ¯uorine as well as using high
¯uorine contents in the inert gas and with elevated reaction
temperatures. In earlier experiments, at least one of these
parameters was set to a notably lower value [4±7] or a more
complex experimental protocol was applied, e.g. by addition
of protic or Lewis acids. Hence, the results of the micro-
reactors indicate that ef®cient temperature control and mass
transfer was exerted, thereby, avoiding hot spots. Any local
increase in temperature would ultimately result in the for-
mation of ¯uorine radicals [4] which will induce side
reactions such as addition or polymerization.
Fig. 4. Selectivity-conversion diagram of direct fluorination of toluene for
the falling film microreactor (FFMR), micro bubble column (MBC (I) and
(II)) and laboratory bubble column (LBC) based on the experiments listed
in Tables 1±3, series A.
the narrow (50 mmÂ50 mm) and wide (100 mmÂ300 mm)
microchannels, and the laboratory bubble column (series A,
Tables 1±3). Nearly constant selectivities are found for the
®rst two microreactors, whereas the corresponding data
decrease with conversion for the micro bubble column using
wide channels and the laboratory bubble column. Remark-
ably, the performance of the falling ®lm microreactor is
clearly superior to that of the laboratory benchmark. In
contrast, the micro bubble column using narrow channels
exceeds the selectivities of the laboratory bubble column
only at high conversion.
The diagram for the micro bubble column equipped with
wide channels is not well understood currently for the
following reasons. Since the corresponding diffusion dis-
tances for mass and heat transport increase when using
larger channels, it was expected that measurements with
this device should result in lower conversions and selectiv-
ities. Surprisingly, the selectivities are higher than for the
micro bubble column with narrow channels when comparing
values obtained at ®xed conversions. For conversion, the
smaller bubble column gives superior performance, even
though the residence times applied were shorter. However,
the origin of the low selectivities remains unclear and
requires more detailed information on hydrodynamics and
transport phenomena.
The same holds for one data point, referring to the highest
conversion, of the micro bubble column using the wide
channels. The corresponding selectivity seems to be unu-
sually low and might be caused by a change of the ¯ow
pattern. Although this information can, for experimental
reasons, not be easily obtained for the ¯uorine/acetonitrile
system, a ¯ow pattern map determined using isopropanol as
liquid indicates a transition between segmented and annular
¯ow for a corresponding set of operating conditions. For
these reasons, it was decided to omit the data record of this
microreactor from the following plots. However, it has to be
emphasized that, concerning the conclusions drawn below,
4.3. Selectivity-conversion diagram
For a detailed analysis of the performance of the reactors,
the representation of the experimental data as a selectivity-
conversion diagram or space-time yield is recommended
[22]. The ®rst type of analysis solely re¯ects the in¯uence of
reaction engineering on the underlying chemistry, whereas
an economic evaluation, in terms of reactor throughput, can
be given in the second case.
In Fig. 4, a selectivity-conversion diagram is given for the
falling ®lm microreactor, the micro bubble column utilizing

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
124
this device gave qualitatively similar results as the other
microreactors when compared to the laboratory benchmark.
micromixing unit into account, giving an idealized reactor
geometry, space-time yield data corrected for the reactor
construction material result (see Fig. 5b). This idealized
reactor geometry, however, represents not the current state-
of-the-art design, but corresponds to a miniaturized proto-
type of future fabrication generation. Data corrected accord-
ing to the overall volume of the current design will be
discussed as well, but are not shown in Fig. 5.
Space-time yields, based on the real reaction volume (see
Fig. 5a) of both microreactors are orders of magnitude
higher than those of the laboratory benchmark. This increase
is mainly caused by much lower residence times achievable
by using these microreactors, the volume corrected yields of
both types of devices being in the same range. While for
completion of the reaction about 3 h are needed in case of
the laboratory bubble column, enhancement of heat and
mass transfer by miniaturization enables ¯uorination to
proceed in a few seconds.
4.4. Space-time yields
To date, a large part of the microreactor volume has to be
designated to achieve ¯uid tightness. In a ®rst step, a
practicable solution was provided by applying ¯at sealing
techniques which inevitably need large areas for build-up of
sealing pressure. The resulting large contribution of parts,
not related to chemical reaction, to the overall reactor
volume is typical for present microreactors, whereas for
conventional reactors this can be neglected. Therefore,
different de®nitions of space-time yields are used in
Fig. 5, based on experiments of series A.
The ®rst approach solely refers to the channel volume, i.e.
the volume where reaction takes place (see Fig. 5a). Due to
the principle of yielding small ¯uid thicknesses, inherent to
microreactors, the reaction channels are separated from each
other by channel walls. Taking this volume and that of the
Even when correcting the data with respect to the
idealized reactor geometry, a signi®cant improvement of
Fig. 5. Comparison of space-time yields of direct fluorination of toluene for the falling film microreactor (FFMR), micro bubble column (MBC (II)) and
laboratory bubble column (LBC) (a) referred to the reaction volume, (b) referred to an idealized reactor geometry.

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
125
performance by using microreactors remains (see Fig. 5b).
Moreover, the current state-of-the-art microreactor design,
far from being optimized regarding system size miniatur-
ization, still has space-time yields, comparable to those of
the laboratory bubble column, when referred to the overall
reactor volume. Concerning safety issues, the use of micro-
reactors allows reduction of the technical expenditure for
laboratory experiments, which handle only small sample
volumes, e.g. for gathering analytical data.
The space-time yield of the microreactors is dependent on
the set of operating conditions chosen. The most important
parameters for an optimization of this ®gure are the ratio of
gaseous to liquid reactant, temperature, solvent, residence
time as well as ¯uorine and substrate concentration. The
in¯uence of the variation of these parameters on target
values such as conversion or selectivity, being proportional
to the space-time yield, is discussed below.
4.5. Variation of fluorine-to-toluene ratio
In Fig. 6a, toluene conversions are given as a function of
the ¯uorine-to-toluene ratio, achieved by variation of ¯uor-
ine volume ¯ow at a ®xed liquid ¯ow rate (series A, Tables 1±
3). The measurements show an increase in conversion both
for the microreactors and the laboratory bubble column. This
indicates that transport hindrances, if present at high ¯uorine
concentrations, do not play a dominant role when using
either devices, e.g. by prohibiting further performance
improvements.
However, there seems to be a small, but distinct difference
between the three types of reactors investigated. This refers
to the amount of ¯uorine contributing to conversion relative
to that introduced in the gas. This ®gure, yet another
measure of reactor ef®ciency, is small at low ¯uorine-to-
toluene ratios for the laboratory benchmark (see Fig. 6b).
For instance, only 6% of the ¯uorine molecules react if a
40% stoichiometric ratio is applied yielding an ef®ciency of
about 15%. In contrast, about 50±75% ef®ciency results for
the two microreactors. For all operating conditions investi-
gated, the ef®ciencies of the microreactors exceed those of
the laboratory bubble column, mostly because of the huge
differences in interfacial areas.
Fig. 6. (a) Conversion of direct fluorination of toluene in the three reactor
types (FFM, MBC (II) and LBC) as a function of the molar ratio of
fluorine-to-toluene. (b) Efficiency of the reactors, defined as conversion
normalized by the molar ratio of fluorine-to-toluene, as a function of the
molar ratio of fluorine-to-toluene.
4.6. Variation of temperature
In the case of the miniaturized and macroscopic bubble
columns, ef®ciency slightly increases when ¯uorine volume
¯ow is enhanced. For the falling ®lm microreactor, however,
performance decreases. This indicates that mass transport
limitations exist at high ¯uorine-to-toluene ratios, i.e. at
high ¯uorine volume ¯ows. Film thickness and interfacial
area being constant, this can only be caused by the de-
crease in residence time of the gas molecules. The time scale
of gas transport to the interface seems to be high compared
to that of transport phenomena in the liquid ®lm. By
reduction of the height of the gas chamber of the falling
®lm microreactor, a simple optimization strategy for
reactor ef®ciency at high ¯uorine-to-toluene ratios is pos-
sible.
The effect of increasing temperature on conversion and
selectivity was investigated for several ¯uorination reactions
[4±7]. As a common result for different aromatic com-
pounds, it was found that, due to increased radical forma-
tion, conversion increases, whereas selectivity decreases. In
the framework of the feasibility experiments presented here
(series A, Tables 1±3), the dependence of these parameters
were analyzed in the temperature range of about 40 to
158C using the falling ®lm microreactor at a ¯uorine-to-
toluene ratio of 40% (see Fig. 7). It turned out that for the
conversions measured the expected relationship was
detected, conversion increased from 15 to 30%. The tem-
perature dependency on selectivity is less clear, showing
scattered data in the range from 33 to 52%. The origin of this

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
126
effect is presently not well understood, and further experi-
mental data are needed.
4.7. Variation of solvent
The choice of solvent mainly affects ¯uorine solubility,
thereby increasing conversion, while its in¯uence on selec-
tivity is more dif®cult to predict. In this context, several
authors described an increase in product yield by increasing
solvent polarity as well as the use of ingredients such as
acids or bases via polarization of the ¯uorine molecule
[5,6,23].
In order to investigate this, ¯uorinations of toluene were
performed in the falling ®lm microreactor and laboratory
bubble column using the protic, polar solvent methanol in
Fig. 7. Conversion (*) and selectivity (~) of direct fluorination of
toluene in the falling film microreactor as a function of temperature.
Fig. 8. Yield as a function of fluorine content in educt gas.

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
127
addition to the aprotic, polar acetonitrile (series A, Tables 1±
3). The results in terms of yields and selectivities for
methanol were worse for both types of reactors. In particular,
the range of selectivities measured in the microreactor is
large and the dependency on the molar ratio of ¯uorine-to-
toluene is not clear. At least, for one set of operating
conditions a relatively high yield of about 14% was found
which also notably exceeds the nearly negligible formation
of mono¯uorinated isomers of only 1% in case of the
laboratory benchmark.
bubble column was demonstrated. Moreover, the reactor
performance, expressed as space-time yield and selectivity-
conversion diagram, is by far superior compared to the
laboratory benchmark. The yields obtained are in the same
order compared to the technically performed Schiemann
process. A further optimization of reactor ef®ciency was
indicated by monitoring the dependence of target quantities,
e.g. conversion and selectivity, as a function of various
operating conditions. It is expected that future experiments
utilizing still higher ¯uorine contents, e.g. up to 100%
¯uorine without any inert gas, at ambient temperatures
may be even superior to technical benchmarks.
4.8. Variation of residence time
The experiments performed for the direct ¯uorination of
toluene in this microreactors are not an entirely systematic
study. To completely survey the ®eld of direct ¯uorination
the following studies have to be performed in future.
Variation of the liquid volume ¯ow for the falling ®lm
microreactor affects the residence time of the ¯owing liquid.
Since the measurements in series Awere limited to only two
values, no general conclusions but only tentative hints can be
given. Increasing the volume ¯ow from 11.6 to 19.6 ml/h has
no effect on toluene conversion, thereby, possibly indicating
that gas diffusion is not limiting for this parameter range (see
Tables 1±4). Selectivity increases slightly which might be
caused by allowing the reaction and, correspondingly, the
heat release to proceed more slowly.
ꢁ The influence of substrate on the performance of direct
fluorination.
ꢁ The solvent used, e.g. polar and non-polar solvents.
ꢁ The substrate concentration, e.g. up to 100%.
ꢁ The fluorine concentration in the gas phase, e.g. up to
100%.
ꢁ The reaction temperature, especially use of room tem-
perature.
ꢁ The use of Lewis acids as a catalyst in direct fluorination.
4.9. Fluorine and substrate concentration
Fluorinations were performed at ¯uorine contents of 10,
25 and 50 vol.% (series B, Table 4). This increase in ¯uorine
concentrations has a remarkable in¯uence on the substitu-
tion pattern which has already been discussed. Moreover, a
nearly linear increase in yield with the ¯uorine content is
found (see Fig. 8).
This increase is mainly due to an enhancement of con-
version, the selectivities being relatively constant (see
Tables 1±4).
In addition to an increase of the concentration of the
gaseous reactant, one experiment was run at a toluene
concentration of 100%, i.e. without any solvent (see
Tables 1±4). Although the yields and conversions obtained
were relatively small, amounting to only a few percentage,
the corresponding selectivities did not decrease compared to
the values obtained at diluted toluene concentrations.
Apart from this quantitative analysis of reactor perfor-
mance, this experiment as well as the ¯uorinations using
high ¯uorine contents demonstrated that in microreactors
even very explosive mixtures can be safely handled. At no
time were explosions or clogging recorded, e.g. prohibiting a
further analysis. In this context, the microreactors proved to
be a useful and precise measuring tool for process regimes,
which are not accessible for conventional laboratory
reactors.
In order to return this technological potential into real
economic bene®t, numbering-up concepts, e.g. a number of
microreactors in parallel for scale-up, have to be developed,
in particular, referring to more compact systems, i.e. devices
with an improved ratio of reaction to construction volume.
Especially, for a signi®cant minimization of the areas
needed for interconnection, the use of irreversible connec-
tion techniques, e.g. laser welding, is a promising approach
for microreactors of the next generation. Special attention
has to be paid for the development of suitable micro¯uidic
structures for manifolding and microstructuring techniques
capable of mass fabrication.
At present, the reaction volume ®lls only a small part of
the overall reactor size, since the residual volume is needed
to guarantee ¯uid tight interconnection. Thus, the current
systems are restricted to laboratory purposes, e.g. for process
development and optimization. Due to the precise setting of
operating conditions and hydrodynamics, e.g. interfacial
areas, accurate data concerning reaction kinetics as well
as mass and heat transport are obtainable, thereby, providing
an excellent basis for simulation and reactor modeling. First
experiments for this purpose are underway.
Acknowledgements
5. Conclusions
The authors thank the German Federal Ministry for
Education and Research (BMBF), the Acordis AG, AXIVA
AG, BASF AG, DuPont GmbH Germany, Merck KGaA and
The feasibility of performing direct ¯uorination using
elemental ¯uorine in the falling ®lm microreactor and micro

È
K. Jahnisch et al. / Journal of Fluorine Chemistry 105 (2000) 117±128
128
[12] J. Shaw, C. Turner, B. Miller, I. Robins, J. Kingston, M. Harper, in
Proceedings of 2nd International Conference on Microreaction
Technology, Topical Conference Preprints, March 9±12, New
Orleans, 1998, p. 267.
the Fonds of the Chemical Industry Germany for ®nancial
support.
È
[13] V. Hessel, W. Ehrfeld, K. Golbig, V. Haverkamp, H. Lowe, M. Storz,
Ch. Wille, Microreaction Technology: Industrial Prospects, in:
Proceedings of 3rd International Conference on Microreaction
Technology, Frankfurt, 18±21 April 1999, Springer, Berlin, 2000,
p. 526.
References
[1] G. Balz, G. Schiemann, Ber. Dtsch. Chem. Ges. 60 (1927) 1186.
[2] G. Schiemann, M. KuÈhnhold, B. Cornils, Liebigs Ann. Chem. 714
(1968) 62.
[14] R.D. Chambers, J. Hutchinson, G. Sandford, J. Fluorine Chem. 100
(1999) 63.
[3] J. Hutchinson, G. Sandford, Organofluorine chemistry, in: R.D.
Chambers (Ed.), Topics in Current Chemistry, 193, 1997, p. 2.
[4] V. Grakauskas, J. Org. Chem. 35 (1970) 723.
[15] R.H. Perry, D.W. Green, Perry's Chemical Engineers' Handbook,
McGraw-Hill, New York, 1997 (Chapters 6±43).
[16] T.C. Thulasidas, M.A. Abraham, R.L. Cerro, Chem. Eng. Sci. 50
(1995) 183.
[5] F. Cacace, A. Wolf, J. Am. Chem. Soc. 100 (1978) 3639.
[6] S.T. Purrington, D.L. Woodard, J. Am. Chem. Soc. 56 (1991) 142.
[7] L. Conte, G.P. Gambaretto, M. Napoli, C. Fraccaro, E. Legnaro, J.
Fluorine Chem. 70 (1995) 175.
È
[17] W.-D. Deckwer, Reaktionstechnik in Blasensaulen, Otto Salle Verlag,
Frankfurt/M., Germany, 1984.
[18] D. Herskowits, V. Herskowits, K. Stephan, Chem. Eng. Sci. 45 (1990)
1281.
[8] W. Ehrfeld, V. Hessel, V. Haverkamp, Microreactors in Ullmann's
Encyclopedia of Industrial Chemistry, 6th Edition, 1999.
[9] H. Kestenbaum, A. Lange de Oliverira, W. Schmidt, F. SchuÈth, W.
[19] O. Levenspiel, Chemical Reaction Engineering, 3rd Edition, 1998,
p. 542.
È
Ehrfeld, K. Gebauer, H. Lowe, Th. Richter, Microreaction Technol-
[20] T.H. Lowry, K. Schueller Richardson, Mechanismen und Theorie in
der Organischen Chemie, Verlag Chemie, Weinheim Germany, 400,
1980, p. 552.
ogy: Industrial Prospects, in: Proceedings of 3rd International
Conference on Microreaction Technology, Frankfurt, 18±21 April
1999, Springer, Berlin, 2000, p. 207.
[21] F. Cacace, P. Giacomello, A.P. Wolf, J. Am. Chem. Soc. 102 (1980)
3511.
[10] R. Srinivasan, I.-M. Hsing, P.E. Berger, AIChE J. 43 (1997) 3059.
È
[11] H. Lowe, W. Ehrfeld, K. Gebauer, K. Golbig, O. Hausner, V.
[22] U.B. McKusick, Ullmann's Encyclopedia of Industrial Chemistry,
5th Edition, Vol. A11, 1998, p. 349.
Haverkamp, V. Hessel, Th. Richter, in: Proceedings of 2nd
International Conference on Microreaction Technology, Topical
Conference Preprints, 9±12 March, New Orleans, 1998, p. 63.
[23] R.D. Chambers, C.J. Skinner, J. Hutchinson, J. Thomson, J. Chem.
Soc., Perkin Trans 1 (1996) 605.