Continuous Platinum-Mediated Hydrogenation of Refametinib Iodo-nitroaniline Key Intermediate DIM-NA: The Combined Challenges of Selectivity and Catalyst Deactivation

Article abstract of DOI:10.1002/ejoc.201700625

The chemoselective, continuous, multistep reduction of iodo-nitroaromatics in a fixed-bed hydrogenation reactor has been investigated. This transformation poses challenges upon both the catalyst and process conditions if high yields are to be achieved. Fi

Full text of DOI:10.1002/ejoc.201700625

DOI: 10.1002/ejoc.201700625
Full Paper
Continuous Hydrogenation | Very Important Paper |
Continuous Platinum-Mediated Hydrogenation of Refametinib
Iodo-nitroaniline Key Intermediate DIM-NA: The Combined
Challenges of Selectivity and Catalyst Deactivation
Todor Baramov,[ [ Jorma Hassfeld, Michael Gottfried, Hermann Bauer,
a,b]
[a]
[c]
[c]
c]
[c]
[a]
Tanja Herrmann, Mourad Ben Said, and Stefan Roggan*
Abstract: The chemoselective, continuous, multistep reduction date. With this catalyst, iodo-nitroaromatic API (active pharma-
of iodo-nitroaromatics in a fixed-bed hydrogenation reactor has ceutical ingredient) intermediate DIM-NA [(2,3-difluoro-5-meth-
been investigated. This transformation poses challenges upon oxy-6-nitrophenyl)(2-fluoro-4-iodophenyl)amine] could be con-
both the catalyst and process conditions if high yields are to verted into the corresponding aniline with high selectivity on a
be achieved. First, the stability and selectivity of four flow-com- preparative scale. UV/Vis spectroscopy was successfully applied
patible Pt catalysts, Pt(X)/C (X = Fe, V), were investigated by for online conversion control, proving to be highly useful for
using 1-iodo-4-nitrobenzene as the model substrate. Granular detecting the effects of catalyst deactivation.
Pt(Fe) on carbon was identified as the most promising candi-
Introduction
should be emphasized that in many cases the use of multipur-
pose batch reactors may represent the best solution. Still, for a
number of specific organic transformations, the advantages of
scaling a synthesis by means of continuous processing time
rather than batch reactor volume or number of batch syntheses
Continuous processing is a long-known technology that has
been successfully applied in the manufacture of large-scale bulk
[
1]
chemicals. Recently, an intense debate has arisen on the ad-
vantages of this type of processing in the pharmaceutical indus-
[3b]
are intriguing.
[
2]
try. Today, the production scales of active pharmaceutical in-
gredients (APIs) often do not reach multi-ton levels, but the
syntheses typically involve multistep transformations of com-
plex and highly functionalized (i.e., precious) substrates in the
liquid phase, which renders selectivity a major challenge. The
advantages of the continuous-processing of fine organic reac-
tions in tubular micro- or meso-reactors, compared with stan-
Heterogeneously catalyzed hydrogenation is certainly a reac-
tion class for which continuous processing should be consid-
ered. When performed in a fixed-bed reactor, continuous-flow
hydrogenation, as compared with batch hydrogenation, bene-
[
5]
fits from improved gas/liquid mixing, heat management, and
a simple and safe utilization of hydrogen at high reaction pres-
[
6]
sures and temperatures. This is especially valid for the
strongly exothermic hydrogenation of organic nitro compounds
[
3]
dard batch reactors, have been extensively reviewed. These
include considerable process intensification in terms of space-
time yield, possible because of extended process windows, and
a higher quality output resulting from a better control of key
process parameters such as reaction (residence) time, tempera-
ture, and, if applicable, pressure. The easier and safer utilization
of highly reactive or toxic reactants and/or intermediates may,
furthermore, facilitate alternative and more efficient synthetic
routes. Thus, a “batch-to-conti” (batch-to-continuous) transfer
r
[7]
(
ΔH = 500–550 kJ/mol), which represents a key fine organic
chemical transformation that is widely employed in industry to
synthesize functionalized primary amines. The reaction pro-
ceeds by a complex, multistep, six-electron reduction pathway
comprising the formation of reactive nitroso and hydroxylamine
[
7]
intermediates, which are prone to a variety of side-reactions.
If halogens are involved, a further selectivity challenge arises,
because the action of H in the presence of a hydrogenation
2
[
4]
often leads to higher productivities or product yields, but it
catalyst can induce hydrodehalogenation (Figure 1, bottom).
Here, the narrower temperature profile resulting from a contin-
uous process may allow for better control of these side-reac-
tions effecting higher selectivities and a uniform quality of the
reactor output. Thus, selectivity represents a major chemical
challenge during nitro-to-amine hydrogenation, and considera-
ble efforts have been devoted to the development of high-
[
a] Bayer AG, Pharmaceuticals,
2096 Wuppertal, Germany
4
E-mail: stefan.roggan@bayer.com
www.bayer.com
[
[
b] Leibniz-Institut für Katalyse e.V. (LIKAT),
Albert-Einstein-Str. 29a, 18059 Rostock, Germany
c] Bayer AG, Engineering & Technology,
[8]
performing catalysts.
5
1368 Leverkusen, Germany
In the present contribution, we report on our studies on the
selective continuous hydrogenation of (2,3-difluoro-5-methoxy-
Supporting information and ORCID(s) from the author(s) for this article are
available on the WWW under https://doi.org/10.1002/ejoc.201700625.
Eur. J. Org. Chem. 2017, 3921–3928
3921
© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Figure 1. Top: Hydrogenation of iodo-nitroaromatic derivative 1 to iodoaniline 2 as part of a possible synthetic route to the anticancer agent Refametinib.
Bottom: Hydrogenation of 1-iodo-4-nitrobenzene (4) to 4-iodoaniline (5). Hydrodehalogenation leads to the formation of the deiodinated products 3 and 6,
lowering selectivity.
6
-nitrophenyl)(2-fluoro-4-iodophenyl)amine (DIM-NA, 1) to the
Building on recent reports by Beller and co-workers on
2
corresponding primary amine 3,4-difluoro-N -(2-fluoro-4-iodo- a novel cobalt catalyst for the selective hydrogenation of
[
8b]
phenyl)-6-methoxybenzene-1,2-diamine (DIM-DAB, 2; Figure 1, (halo-)nitroaromatics,
top) on a preparative scale. Compounds 1 and 2 are key inter- glomerates with particle sizes of 200–800 μm [d(50) = 564 μm]
mediates in the multistep synthesis of the antitumor agent as the support material for related cobalt catalyst,
Refametinib, a highly potent and selective MEK1/2 inhibitor, CoNGr@CNT, with a morphology more suitable for applications
which is currently undergoing phase II clinical trials for late- in flow chemistry. This catalyst proved to be highly selective,
stage cancer treatment.
we used carbon nanotube (CNT) ag-
a
[
12]
[
9]
and under optimized conditions (80 °C, 60 bar) 1-iodo-4-nitro-
Regarding hydrodehalogenation, the highest lability has benzene was hydrogenated without any notable deiodination.
[
7]
been reported for the iodine in the amine product. This However, the productivity of the catalyst [3.0 mmol(5)/
means that the dehalogenation side-product primarily forms in {g(cat.) h}] was rather unsatisfactory due to a comparatively low
a follow-up reaction of the targeted product (Figure 1, bottom), catalyst activity and the low bulk density (0.27 g/mL) of the
[
12]
so iodo-nitroaromatics can be considered the most challenging material.
substrates for selective hydrogenation up to full conversion of
Doped Pt catalysts, 1 % Pt(X)/C (X = V, Fe), can also be used
[
10]
the starting nitro compound. A selective catalyst must, there- for the selective hydrogenation of (halo)nitroaromatics. Their
fore, be active enough to suppress the accumulation of reactive commercial, standardized availability is an advantage, at least
intermediates from nitro group reduction, but, at the same for industrial applications, especially under the constraints of
time, it should not show a pronounced activity towards de- GMP (good manufacturing practice) manufacture. A 1 % Pt(V)/
iodination upon reaching full conversion.
C catalyst has been shown to perform reasonably well in the
To date, only a few heterogeneous catalysts have been ap- conti-hydrogenation of 4.[10] However, mechanical fragility and
plied in the continuous hydrogenation of halo-nitroaromatics. severe deactivation have prohibited the repeated use of this
[
10,12]
Loos et al. reported that Raney cobalt could be a selective cata- catalyst, even in small-scale processes.
We, therefore, de-
lyst for this transformation. They found that conti-hydrogen- cided to evaluate in more detail the potential of this family of
ation of the highly labile 4 (Figure 1) in a 5 mL fixed-bed reactor catalysts for the conti-hydrogenation of DIM-NA, not
at 80 °C and 85 bar gave only 1.5–2 % of the deiodinated side- only regarding the selectivity but also the scalability of the
[
10]
product aniline at full conversion.
Encouraged by those process up to preparative amounts useful for early preclinical
promising results, we recently scaled the conti-hydrogenation pharmacological studies.
process over a fixed bed of Raney Co catalyst to 50 mL reactor
volume. Using a 9 mm (inner diameter, ID) reactor tube, we
successfully converted 1 kg of Refametinib intermediate DIM-
Results and Discussion
[
11]
NA (1) into DIM-DAB (2) at 60 °C within 8 h (Figure 1).
The
selectivities obtained were remarkable: At >99.5 % conversion, When switching from batch hydrogenation in an autoclave to
only negligible amounts of deiodinated 3 (≤0.5 %) or other a continuous-flow process in which the catalyst is employed in
side-products were formed throughout the process, and the a packed bed, special attention has to be given to the morphol-
industrial specifications for this step were clearly matched. ogy of the catalyst. For proper flow dynamics, the particle size
However, high pressure drops (up to 80 bar) resulted from the should match the dimensions of the reactor tube to avoid non-
small particle size (40 μm) of the Raney Co catalyst and the uniform process conditions throughout the reactor due to pres-
throughput employed.
sure drops or channeling. Accordingly, suitable ratios between
Eur. J. Org. Chem. 2017, 3921–3928
www.eurjoc.org
3922 © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
the diameter of the reactor tube (D) and the catalyst particle
[
13]
size (d ) have been reported to be D/d > 10.
p
p
Vanadium- and iron-doped carbon-supported platinum cata-
lysts as well as undoped Pt/C with a morphology suitable for
application in our flow reactor were obtained from two com-
mercial catalyst providers as 0.8 mm catalyst extrudates and
0.4–0.8 mm granules, respectively (Table 1 and Figure S1 in the
Supporting Information). To determine the best candidate for
the envisioned preparative reactions in a 50 mL reactor employ-
ing 1 as the starting material, the stability and selectivity of the
five catalysts were evaluated using 1-iodo-4-nitrobenzene (4) as
a model substrate (Figure 1).
Table 1. Catalysts investigated for their stability and selectivity in the continu-
ous hydrogenation of 4 at 0.43 mmol(4)/{g(cat.) min}. c(4) = 0.1
M
in THF/
Figure 2. Deactivation profiles of the five carbon-supported Pt catalysts em-
ployed in this study during the selective continuous hydrogenation of 4-
iodonitrobenzene (4; see Figure S3 and Table S1 in the Supporting Informa-
tion for the dehalogenation profiles).
H
2
O (95:5, v/v), T = 80 °C, p = 40 bar, 6 equiv. H
2
.
Catalyst[a]
Amount[b]
g]
Flow rate
[mL/min]
Deactivation[c]
[%]
Aniline
[%][
d]
[
gran
1
1
1
1
% Pt(V)/C
% Pt(V)/C
% Pt(Fe)/C
% Pt(Fe)/C
0.79
0.93
0.84
1.17
0.79
3.4
4.0
3.6
5.0
3.4
16.5
19.8
10.6
24.5
24.6
1.2
1.9
1.4
4.6
9.0
extr
gran
[e]
sion. For a conti-process based on a fixed-bed reactor, three
cases can be considered that may influence the processing op-
tions and selectivity of the reaction (Figure 3).
extr
extr
1
% Pt/C
[
[
a] The catalyst morphology was either granules (gran) or extrudates (extr).
b] Amount of catalyst required for a 2.75 mL catalyst bed, the remaining
volume of the catalyst cartridge was filled with inert material (see the Sup-
porting Information). [c] After 12 h time on stream with respect to the initial
activity, as calculated from the GC area%. [d] Compared at around 78–81 %
conversion; see Table S1 in the Supporting Information. [e] Reported at 87 %
conversion as lower conversions were not reached within the experimental
time.
A 0.1
M
solution of 4 was hydrogenated at 80–99 % conver-
sion in a commercial cartridge reactor (10 mm ID, see the Sup-
porting Information for details) loaded with 2.75 mL of the re-
spective catalyst (Table 1). For proper evaluation and compari-
son of the different materials, the flow rate was adjusted to
Figure 3. Different scenarios for the selective continuous hydrogenation of
halo-nitroarenes in a fixed-bed reactor. Case I (left): stable and highly select-
ive catalyst. Case II (middle): combination of a deactivating but selective cata-
lyst. Case III (right): deactivating catalyst with modest to low selectivity re-
garding the follow-up dehalogenation reaction.
achieve
a uniform workload [mmol(4)/{g(cat.) min}]. The
hydrogenation experiments were conducted over a total period
of 12 h with an interruption after 5 h, during which the catalyst
bed was stored in THF/H O (95:5, v/v) overnight. The product
2
mixtures were analyzed at regular intervals by gas chromatog-
Case I (Figure 3, left) represents the favorable situation in
raphy (GC) for conversion of the starting material and the con- which a non-deactivating and highly selective catalyst is avail-
tent of the deiodinated product aniline.
able. Under these conditions, neither catalyst deactivation af-
For all catalysts, a steady deactivation profile was found in fects the targeted conversion over time nor will selectivity dete-
the investigated timespan (Figure 2), which may be due to riorate by the presence of the excess catalyst bed fostering the
metal leaching or inhibition of the Pt catalyst by the product follow-up dehalogenation reaction. This constellation allows for
[
14]
amine. After 12 h time on stream (TOS), the extent of deacti- variation of the reactor throughput within a given range, and
vation was in the range of 11–25 % (Table 1), and there were the process-optimization efforts are comparatively low. In
no indications that the catalyst activity would converge towards case II (Figure 3, middle), a deactivating but still highly selective
a stable value. Pt(Fe)/Cgran turned out to be the best-performing catalyst is present. Here, the implementation of an excess cata-
catalyst, showing the highest stability as well as the highest lyst bed (as realized by a proper flow rate) is a possible measure
activity and selectivity (Figure 2, Table 1). As expected, the se- to compensate for catalyst deactivation because the presence
extr
lectivity of the undoped catalyst Pt/C
was rather poor with of excess catalyst will not affect the selectivity of the process.
a 4.7-fold higher amount of aniline in the product mixture (at Case III (Figure 3, right) represents the most challenging situa-
0 % conversion) as compared, for example, with the 1 % Pt(V)/ tion. It illustrates the combination of a deactivating catalyst and
8
C
extr
catalyst (see Figure S3 and Table S1).
a reaction system characterized by the existence of a follow-up
The combination of catalyst deactivation and a follow-up reaction of the targeted product(s) catalyzed by the same mate-
dehalogenation side-reaction poses considerable challenges to rial (hydrodehalogenation in the present case). Here, excess
the task of selective continuous hydrogenation at full conver- catalyst bed will (further) increase the concentration of such
Eur. J. Org. Chem. 2017, 3921–3928
www.eurjoc.org
3923
© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
byproducts, and to minimize their formation (1) the catalyst (1) hydrogenation on a preparative scale using the reactor
[
11]
must have a minimum selectivity to inhibit the follow-up reac- setup schematically shown in Figure 4 (see ref. and the Sup-
tion at lower conversions of the reactant and (2) the process porting Information for a detailed description of the experimen-
conditions have to be adjusted to exactly reach full conversion tal setup). In a typical experiment, the liquid reactant solution
just at the outlet of the reactor. In such a scenario, if the catalyst and gaseous hydrogen are combined in a microstructured gas/
employed suffers from deactivation, the conversion will quickly liquid mixer. The resulting reaction mixture is then passed over
drop below the tolerable limit and immediate measures will be a heated catalyst bed packed into a 50 mL reactor tube with
required to avoid product contamination by too much unre- 10.2 mm ID. The reaction pressure is established by a back-
acted starting material. The latter can be simply accomplished pressure retention valve installed following the reactor outlet.
by reducing the reactant's flow rate, that is, by increasing the On exiting the pressurized part of the setup, the product stream
residence time, but this requires a fast-responding tool for on- is collected in either a nitrogen-purged product or waste con-
line conversion determination and is at the cost of the through- tainer. Two valves (V16 and V17 in Figure 4) installed after the
put of the process.
back-pressure regulator allow for incorporation of an online
In view of the above discussion, if there is no extraordinarily monitoring module prior to product collection.
selective catalyst available, the conti-hydrogenation of iodo-
nitroarenes is to be classified in category three.
Table 2 summarizes the results obtained for the continuous-
flow hydrogenation of DIM-NA over Pt(Fe)/Cgran.
The experiments were performed with 10.7 g of the iron-
doped Pt catalyst loaded into the heated part of the reactor
tube giving a catalyst bed volume of 30 mL. Entries 1–3
Continuous DIM-NA Hydrogenation
Due to its superior activity and selectivity among the catalysts (Table 2) show the results of a 100 min experimental run during
gran
evaluated, we chose Pt(Fe)/C
for further studies of DIM-NA which the process temperature and flow rate were varied. The
Figure 4. Reactor setup for continuous hydrogenation (see the Supporting Information, Figure S4 for a representation including a picture).
Eur. J. Org. Chem. 2017, 3921–3928 www.eurjoc.org © 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
3924

Full Paper
Table 2. Summary of the results obtained for the continuous hydrogenation of DIM-NA over the 1 % Pt(Fe)/Cgran catalyst.[a]
Entry
Flow rate
mL/min]
Catalyst workload
[mmol(1)/{g(cat.) min}]
p
[bar]
T
[°C]
X(1)[b]
[%]
S(2)[b]
[%]
S(3)[b]
[%]
S(other)[b]
P(2)[c]
[g(2)/h]
[
[%]
[
[
[
d]
d]
d]
40–43[e]
60
60
60
80
1
2
3
4
5
10
10
12
0.094
0.094
0.112
0.187
0.187
40
40
40
60
60
99.1
>99.9
99.6
99.7
99.6
99.0
98.7
98.8
99.4
98.9
0.6
0.9
0.8
0.5
1.1
0.4
0.5
0.4
0.0
0.0
23.2
23.3
27.9
46.9
46.6
[
f]
f]
10
10
[
[
a] 10.7 g of 1 % Pt(Fe)/Cgran granules loaded into the reactor tube gave a catalyst bed volume of 30 mL. [b] Conversion (X) and selectivity (S) were derived
from the HPLC area% (see the Supporting Information). [c] Productivity (P). [d] A 0.1 solution of 1 in THF/H O (95:5, v/v) was processed, 6.0 equiv. H were
applied. [e] The temperature of the heating fluid increased slightly during the experiment. [f] A 0.2 solution of 1 was processed, 3.4 equiv. H were applied.
M
2
2
M
2
compositions of the product mixtures were determined by only S(3) = 0.5 % dehalogenated side-product (Table 2, Entry 4).
HPLC analysis of samples taken at regular time intervals after No further side-products were detected by HPLC (see Fig-
reaching stationary reactor output for a given set of process ure S8). This reactor output corresponds to a productivity P(2) =
parameters (ensured by unchanged compositions of sequential 46.9 g/h, which was confirmed by analysis of the crude product
product samples, see the Exp. Sect.). Passing a 0.1
M DIM-NA (approx. 10 g) isolated after a collection period of about 15 min
solution in THF/H O (95:5, v:v) at 40 °C and 40 bar at 10 mL/ (see Figure S9). The solid material was found to contain minor
2
min over the catalyst bed in the presence of excess H2 gas amounts of Pt (6 ppm; ICP-OES), which indicates some Pt leach-
(
Table 2, Entry 1) led to 99.1 % conversion. No pressure drops ing, but overall it can be considered a rather small amount for
[
15]
were observed at the applied liquid and gas flow rates on com- an isolated crude solid material.
Surprisingly, increasing the
paring the reactor inlet pressure with the applied back pressure. process temperature to 80 °C did not lead to full conversion
The selectivity of the reaction was excellent, with only 0.6 % of 1 (Table 2, Entry 5). Instead, DIM-NA conversion remained
deiodinated side-product 3 and negligible amounts (0.4 %) of unchanged (99.6 %), and there was a notably higher degree
other byproducts in the product solution (see Figure S5 in the of deiodination [S(3) = 1.1 %]. Greater dehalogenation can be
Supporting Information). Deiodination was, therefore, consider- expected from higher reaction temperatures at a given conver-
ably lower than in the case of the hydrogenation of 4 (5.3 % sion. The lack of increase of DIM-NA conversion at higher tem-
aniline at 97.8 % conversion; Figure 2), but that reaction was perature may be an effect of catalyst deactivation hitherto
performed at a higher temperature and throughput. The pro- unnoted due to comparatively short experimental runs and col-
ductivity of DIM-DAB was P = 23.2 g(2)/h under these condi- lection times. However, it matches observations made during
tions (Table 2, Entry 1), as calculated from the concentration, catalyst screening experiments with 1-iodo-4-nitrobenzene (Fig-
flow rate, conversion, and selectivity. Accordingly, running a ure 2).
DIM-NA hydrogenation process according to Entry 1 (Table 2)
allows for the simple synthesis of several hundred grams of raw
DIM-DAB in an ordinary laboratory hood within a couple of
Online Monitoring
hours. The low residual amounts of starting material and side-
[
11]
We Although the Pt(Fe)/Cgran catalyst exhibits very good selectivity
products enable an easy purification by crystallization.
isolated 4.9 g of raw material by solvent evaporation after a for DIM-NA hydrogenation at high conversions, it suffers from
collection period of about 15 min, and HPLC analysis gave the deactivation. As discussed above, processing a reaction that is
same composition as for the reactor sample [i.e., X = 99.3 %, characterized by the combination of a selectivity challenge due
S(2) = 99.5 %, S(3) = 0.5 %], thereby confirming a stable and to a follow-up reaction, and a deactivating catalyst requires a
stationary reactor output during this period (see Figure S6). In- quick-responding, efficient, online-monitoring tool to recognize
creasing the process temperature to 60 °C led to full DIM-NA conversions dropping below a predetermined limit. In such
conversion without any detectable amounts of 1 in the product cases, the residence time can be adapted to reconstitute the
solution (Table 2, Entry 2; see Figure S7). Deiodination was targeted conversion without interrupting the entire conti-proc-
slightly higher (0.9 %) under these conditions, which is a conse- ess. In the present case, up to 2 % of starting material can be
quence of both the higher temperature and some excess cata- tolerated in the raw product to minimize the formation of im-
lyst bed, which according to the discussion above, in the pres- purity 3. At the same time, online monitoring has to ensure
ence of excess H triggers the dehalogenation follow-up reac- that the conversion remains just at (or closely below) 100 %
2
tion. At higher throughput, that is, upon increasing the flow to limit the occurrence of undesired overhydrogenation. The
rate to 12 mL/min but maintaining constant the reactor pres- pronounced color difference between the starting nitro com-
sure and temperature (Table 2, Entry 3), conversion and de- pound 1 (yellow) and the amino product 2 (colorless; see Fig-
iodination dropped to 99.6 % and S(3) = 0.8 %, respectively, ure S10 in the Supporting Information) makes UV/Vis spectro-
which is in accord with expectations. Under these conditions, scopy a suitable tool for online-monitoring of the conversion of
P(2) increased to 27.9 g/h. To further increase the productivity the reaction. A UV/Vis detection module can be easily inte-
of the process, a higher-concentrated reactant solution was em- grated into a continuous product stream, allowing for fast anal-
ployed. At 10 mL/min, 60 bar, and 60 °C, a 0.2
M DIM-NA reac- ysis at specific wavelengths. Accordingly, the experimental
tant solution was hydrogenated with 99.7 % conversion and setup was extended by incorporating a UV/Vis flow cell and
Eur. J. Org. Chem. 2017, 3921–3928
www.eurjoc.org
3925
© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
detector into the depressurized part prior to product collection
Figure 4 and Figure S4). Because the presence of hydrogen gas
(
affects UV detection, the product stream was passed through
an inline gas/liquid phase separator before entering the flow
[
11]
cell (see Figure S4). On comparing the full-range UV/Vis spec-
tra of diluted and more concentrated (0.1, 0.2 ) solutions of
M
starting material 1 and product 2, we selected 425 nm as a
suitable wavelength to monitor the concentration of (residual)
DIM-NA in the product stream (see Figure S11). The UV/Vis ab-
sorption was calibrated externally with model mixtures of 1 and
2
in THF/H O, mimicking conversions in the range of 95–100 %
2
at the concentration level of interest (see Figure S12), to allow
for immediate determination of the actual conversion from the
absorption recorded online. The functionality of the new on-
line-monitoring module was checked within a conti-hydrogen-
ation experiment conducted in the 50 mL reactor by pumping
Figure 5. Online conversion control of the DIM-NA hydrogenation by UV/Vis
spectroscopy. HPLC analytical results obtained from individual samples taken
during the process are superimposed on the respective TOS (blue marks;
Table 4).
a 0.1
M
DIM-NA (1) solution at various flow rates through the
catalyst bed (Table 3).
ferent times during the experiment. Although the results from
online analysis and offline HPLC were generally in good agree-
ment (Table 4), the conversions determined by HPLC analysis
were found to be systematically lower than those derived from
online UV absorption (Figure 5, Table 4), with slightly higher
discrepancies (≤1.0 %) at around 95 % conversion than at 99 %
conversion (≤0.3 %). To further improve the agreement be-
tween the two techniques, calibration of the UV/Vis online ab-
sorption system should be performed with the corresponding
offline HPLC analyses rather than by using model sample mix-
tures. Catalyst deactivation reached 1.8 % over the entire exper-
imental runtime (165 min) at an average catalyst workload of
Table 3. Process conditions applied in the continuous hydrogenation of DIM-
NA (1) with online conversion control. c(1) = 0.1
2
M in THF/H O (95:5, v/v), T =
4
0 °C, p = 40 bar, 3.4 equiv. H (6.1, 7.6, 9.1 L/h).
2
TOS[a]
Flow rate[b]
Catalyst workload
[
min]
–67
7–121
21–165
[mL/min]
[mmol(1)/{g(cat.) min}]
1
12
15
18
0.087
0.108
0.130
6
1
[a] Time on stream. [b] Liquid flow rate (DIM-NA solution).
The reactor setup was first purged with pure solvent, and
the experiment was subsequently initiated (t = 0 in Figure 5)
by applying a liquid flow rate of 12 mL/min. After an equilibra-
tion period of approximately 12 min, full DIM-NA conversion
0
.11 mmol(1)/{g(cat.) min}. This compares well with the obser-
gran
vations made for the Pt(Fe)/C
catalyst in the catalyst screen-
ing pre-study performed on 1-iodo-4-nitrobenzene (Table 1,
Figure 2). In that case, 2.2 % deactivation occurred at
(
Figure 5) was achieved (as indicated by the online UV/Vis ab-
sorption values; see Figure S13 in the Supporting Information).
However, within the following 55 min, the conversion steadily
dropped to 99.5 % (revealed by a continuously increasing UV/
Vis absorption). This result points to a slow but ongoing catalyst
deactivation, which, as in the case of the hydrogenation of 4
0
.43 mmol(4)/{g(cat.) min} within 180 min.
Table 4. Comparison of DIM-NA conversions determined by online UV/Vis
monitoring and offline HPLC analysis.
25
X(UV425
[%]
TOS
min]
Flow rate
[mL/min]
UV4
)
X(HPLC)
[%]
(
(
Figure 2), may result from the combined effects of Pt leaching
see above) and product inhibition due to adsorption of the
[
[a.u.][
a]
[
14,16]
45
12
12
12
15
15
15
18
18
18
0.058
0.068
0.074
0.199
0.211
0.216
0.381
0.394
0.397
99.7
99.5
99.5
97.9
97.7
97.7
95.6
95.4
95.4
99.4
99.3
99.2
97.3
97.2
97.1
94.6
94.5
94.4
product amine at the metal surface.
The flow rate was then
5
6
7
7
raised to 15 mL/min, and, as expected, the DIM-NA conversion
decreased as a consequence of the reduced residence time of
the reactants over the catalyst bed. As can be seen in Figure 5
103
112
121
151
159
165
(
and Figure S13), the reactor output quickly adapted to the
UV
flow-rate change, that is, the drop in conversion from X
9
=
UV
9.5 % (t = 69 min) to X = 98.4 % (t = 73 min) occurred within
only 4 min. Ongoing catalyst deactivation is reflected by the
UV/Vis absorption profile recorded over the entire 54 min dur-
[
a] Absorption at 425 nm, see Figure S13 in the Supporting Information.
ing which the process was run at 15 mL/min; the conversion
The online-monitoring experiment clearly shows that varia-
UV
ultimately dropped to X = 97.7 % (at t = 121 min). Eventually, tions in the DIM-NA conversion (as induced by changes in flow
the liquid flow rate was increased to 18 mL/min, and again, the rate and catalyst deactivation in the present case) can be easily
resulting change in conversion as well as the ongoing catalyst and quickly detected in situ by UV/Vis spectroscopy. Thus, if as
deactivation are nicely reflected by the online UV/Vis absorp- a consequence of catalyst deactivation the conversion drops
tion profile (Figures 5 and S13).
below a specific limit, the targeted range can be easily read-
The accuracy of the online UV/Vis conversion determination justed by adapting the flow rate to the required UV/Vis absorp-
was verified by HPLC analysis of product samples taken at dif- tion.
Eur. J. Org. Chem. 2017, 3921–3928
www.eurjoc.org
3926
© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
Conclusions
Hydrogenation was performed at 40 bar, 80 °C, and flow rates indi-
vidually adapted to the amount of the respective catalyst for a con-
sistent catalyst workload of 0.43 mmol(4)/{g(cat.) min} (Table 1). At
We have shown that hydrogenation of the iodo-nitroaromatic
API precursor DIM-NA (1) to the corresponding aniline DIM-DAB the end of the experiment, the system pressure was released, heat-
2) can be successfully achieved in the presence of 1 % Pt(Fe)/C ing switched off, and the reactor purged with THF/H O (at 2 mL/
(
2
in a continuous process on a preparative scale. The pronounced min) while cooling to room temperature for 1 h.
selectivity challenge of this reaction, arising both from the step- Procedure for the Continuous Hydrogenation of DIM-NA: The
wise, six-electron reduction of the nitro functional group and continuous hydrogenation of DIM-NA was performed in a 50 mL
a follow-up deiodination reaction of the product, poses high reactor tube (L = 60.9 cm, ID = 1.02 cm) installed in the experimen-
tal setup shown in Figure 4 and described in the Supporting Infor-
demands on the catalyst and process conditions if high product
yields are to be achieved. Five Pt catalysts with different do-
pants and morphologies of the carbon support (granules or
extrudates) were evaluated by employing a fixed-bed reactor.
We first assessed their performance in the conti-hydrogenation
mation (Figure S4). The heated part of the reactor tube was loaded
gran
with 10.7 g of fresh 1 % Pt(Fe)/C
resulting in a catalyst bed with
a volume of 29.9 mL (L = 36.6 cm). The reactor volume not covered
by the heating medium was filled with glass beads. Before starting
the hydrogenation, the reactor was purged for 10 min with a mix-
of iodo-nitroaromatics by using 1-iodo-4-nitrobenzene (4) as
ture of THF/H O (10 mL/min) and N2 (25 L/h) keeping the back-
2
gran
the substrate. In a 5 mL flow reactor, 1 % Pt(Fe)/C
was identi-
pressure valve at atmospheric pressure while heating to the corre-
fied as the best catalyst both in terms of selectivity and stability. sponding reaction temperature (Table 2; the reaction temperature
Subsequently, we used this material for the successful conti- was typically adjusted at thermocouple T08, see Figure 4 and Fig-
hydrogenation of DIM-NA on a 50 mL reactor scale. Excellent ure S4). After reaching the desired temperature, the nitrogen flow
was replaced by H (10 L/h), and the back pressure was gradually
selectivities at >99 % conversion and productivities of up to
7 g(DIM-DAB)/h were achieved by employing 0.1 or 0.2 reac-
2
increased to the desired value (Table 2) maintaining the liquid flow
rate (10 mL/min) for additional 10 min. After catalyst activation un-
der the reaction conditions, the liquid feed was switched to the
4
M
tant solutions in THF/H O. Online UV/Vis spectroscopy proved
2
to be a useful tool for conversion control, allowing for in situ
compensation of the effects of catalyst deactivation by flow-
rate adaptation.
reactant solution (0.1 or 0.2
M DIM-NA in THF/H O, 95:5, v/v), and
2
the H flow was adjusted accordingly (Table 2). The composition of
2
the reaction mixture was analyzed by HPLC (see the Supporting
Information for details) in regular intervals to verify the steady state
The combined challenges of catalyst deactivation and select-
ivity resulting from the presence of a side-reaction following of the reactor output. The same procedure was repeated for each
product formation during continuous processing were evalu- set of reaction parameters. The product solution was collected in a
nitrogen-purged phase segregator from which the sample solution
was removed for further isolation if desired (see below). At the end
ated in depth and the consequences analyzed for three differ-
ent scenarios.
of the experiment, the reactor was purged with THF/H O (10 mL/
2
min)/N (25 L/h) and the back pressure was gradually released to
2
atmospheric pressure while cooling to room temperature. The prod-
uct solutions obtained from process conditions E1 and E4 (Table 2)
were collected during approximately 15 min. Solvent removal under
Experimental Section
General: DIM-NA (1), DIM-DAB (2), and deiodo-DIM-DAB (3) were
provided by Bayer AG and used as received. 1-Iodo-4-nitrobenzene reduced pressure yielded 4.9 (81 %) and 10.2 g (86 %) of a grayish
(4; Sigma–Aldrich, Lot STBD1156V), 4-iodoaniline (5; TCI, Lot crude product, respectively (see Figures S6 and S9 in the Support-
KCWYG-RR), and aniline (6; Sigma–Aldrich, Lot 10420MD) were pur- ing Information). Yield losses were due to sampling from the prod-
gran
gran
extr
chased from commercial suppliers. Pt(Fe)/C
and Pt(V)/C
were
were
uct stream during the collection period. Analytical data for DIM-
extr
extr
[9e] 1
obtained from Heraeus. Pt(Fe)/C , Pt(V)/C , and Pt/C
DAB (2):
H NMR (400 MHz, CDCl ): δ = 7.37 (dd, J = 10.6, 1.9 Hz),
3
provided by Chimet. The solvents (THF: Acros, 99.6 %, stabilized 7.22 (dt, J = 8.5, 1.3 Hz, 1 H), 6.61 (dd, J = 11.6, 7.0 Hz, 1 H), 6.23
with butylated hydroxytoluene, BHT; acetonitrile: LC-MS Chroma- (dt, J = 9.1, 0.8 Hz, 1 H), 5.34 (s, 1 H), 3.85 (s, 3 H) ppm. C NMR
solv) were purchased from commercial sources. H (99.999 %) was
1
3
2
(101 MHz, CDCl ): δ = 151.69 (d, J = 245.8 Hz), 142.46 (d, J =
3
purchased from Air Liquide. All materials were used without further
purification. H PO and KH PO were used for buffer preparation.
Deionized water, used for the preparation of the reactant solution
and the HPLC buffer solution, was purified by employing a Millipore
Milli-Q® system.
12.0 Hz), 142.44 (dd, J = 237.7, 13.0 Hz), 141.23 (d, J = 228.9,
13.4 Hz), 133.46 (d, J = 3.4 Hz), 133.01 (d, J = 10.9 Hz), 130.49 (s),
123.87 (d, J = 20.8 Hz), 116.36 (d, J = 2.9 Hz), 114.38 (d, J = 14.6 Hz),
3
4
2
4
1
9
98.17 (d, J = 22.4 Hz), 78.75 (d, J = 7.7 Hz), 56.09 (s) ppm. F NMR
(376 MHz, CDCl ): δ = –132.15 (s, 1 F), –150.46 (d, J = 22.9 Hz, 1 F),
3
–
3
157.13 (d, J = 23.9 Hz, 1 F) ppm. HRMS (ESI): calcd. for C H F IN O
94.9868 [M + H] ; found 394.9882.
13 11 3 2
Procedure for Catalyst Stability Tests: The catalyst cartridge of a
commercial Ehrfeld cartridge reactor 240 was loaded as follows (see
Figure S2 in the Supporting Information): glass wool (0.150 g), the
+
Procedure for the Continuous Hydrogenation of DIM-NA with
Online Monitoring: DIM-NA hydrogenation was carried out by us-
, Pt/C ], as indicated in Table 1, glass wool (0.150 g), glass ing the experimental setup shown in Figure 4, extended to UV/Vis
beads 0.75–1.0 mm (ca. 2.4 g), glass wool (0.150 g), giving a catalyst online monitoring (see Figure S4 in the Supporting Information). A
bed volume of 2.75 mL (3.5 × 1.0 cm). The desired reaction condi- 50 mL reactor tube (L = 79.0 cm, ID = 0.9 cm) was loaded with
gran
gran
extr
corresponding catalyst [Pt(Fe)/C , Pt(V)/C , Pt(Fe)/C , Pt(V)/
extr
extr
C
tions (T, p, liquid flow rate, H flow rate) were established by using
13.8 g of fresh catalyst giving a catalyst bed volume of 44.3 mL (L =
69.7). A 0.1 DIM-NA solution in THF/H O (95:5, v:v) was hydrogen-
2
THF/H O (95:5, v/v) as the solvent. After catalyst activation for
M
2
2
3
0 min under these conditions, a 0.1
M
solution of 1-iodo-4-nitro-
ated under the conditions given in Table 3. The conversion of the
staring material was monitored online based on the UV absorption
benzene (4) in THF/H O was passed over the catalyst bed at the
2
desired flow rate for a total of 12 h with an interruption after 5 h, recorded at 425 nm as well as offline by HPLC (see the Supporting
during which the catalyst was stored in THF/H O overnight. Information for details).
2
Eur. J. Org. Chem. 2017, 3921–3928
www.eurjoc.org
3927
© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim

Full Paper
[
6] a) M. Irfan, T. N. Glasnov, C. O. Kappe, ChemSusChem 2011, 4, 300–316;
b) K. Mennecke, R. Cecilia, T. N. Glasnov, S. Gruhl, C. Vogt, A. Feldhoff,
M. A. Larrubia Vargas, C. O. Kappe, U. Kunz, A. Kirschning, Adv. Synth.
Catal. 2008, 350, 717–730.
Acknowledgments
This research was supported by the Innovative Medicines Initia-
tive (IMI) joint undertaking project CHEM21, under grant agree-
ment no. 115360, with financial contributions from the Euro-
pean Union's Seventh Framework Program (FP7/2007-2013) and
the European Federation of Pharmaceutical Industries and Asso-
ciations (EFPIA). We wish to thank K. Petry (Bayer AG) for excel-
lent support during experimentation as well as R. Pellegrini
[
[
7] M. Pietrowski, Curr. Org. Synth. 2012, 9, 470–487.
8] a) X. Wang, Y. Li, J. Mol. Catal. A 2016, 420, 56–65; b) F. A. Westerhaus,
R. V. Jagadeesh, G. Wienhöfer, M.-M. Pohl, J. Radnik, A.-E. Surkus, J. Rab-
eah, K. Junge, H. Junge, M. Nielsen, A. Brückner, M. Beller, Nat. Chem.
2013, 5, 537–543; c) Y. Motoyama, Y. Lee, K. Tsuji, S. H. Yoon, I. Mochida,
H. Nagashima, ChemCatChem 2011, 3, 1578–1581; d) A. Corma, P. Serna,
P. Concepción, J. J. Calvino, J. Am. Chem. Soc. 2008, 130, 8748–8753;
e) M. Takasaki, Y. Motoyama, K. Higashi, S.-H. Yoon, I. Mochida, H. Naga-
shima, Org. Lett. 2008, 10, 1601–1604; f) G. Wu, M. Huang, M. Richards,
M. Poirier, X. Wen, R. W. Draper, Synthesis 2003, 1657–1660; g) H.-U. Bla-
ser, H. Steiner, M. Studer, ChemCatChem 2009, 1, 210–221; h) U. Siegrist,
P. Baumeister, H.-U. Blaser in Catalysis of Organic Reactions (Ed.: F. Herkes),
Marcel Dekker, New York, 1998, pp. 207–219.
(Chimet) and B. Geiger (Heraeus) for providing the catalyst sam-
ples used in this study.
Keywords: Flow chemistry · Hydrogenation · Heterogeneous
catalysis · Nitroarenes · Online monitoring · UV/Vis
spectroscopy
[
9] a) P. Fey, A. C. Mayer (Bayer AG), patent application, WO 2012/163799 A1,
2012; b) D. Xiao, L. Zhu, S. Wang, Z. Liang, W. Hu (Chemizon, Ltd.), patent
application, WO 2010/145197 A1, 2010; c) A. Maderna, J. M. Vernier (Ar-
dea Biosciences, Inc.), patent application, WO 2011/009541 A1, 2011;
d) C. Iverson, G. Larson, C. Lai, L.-T. Yeh, C. Dadson, P. Weingarten, T. Ap-
pleby, T. Vo, A. Maderna, J.-M. Vernier, R. Hamatake, J. N. Miner, B. Quart,
Cancer Res. 2009, 69, 6839–6847; e) J. M. Vernier, C. E. Rowlings, J.-L.
Girardet, S. Dimoc, B. Quart, J. N. Miner (Ardea Biosciences, Inc.), patent
application, US 8648116 B2; f) https://clinicaltrials.gov/ct2/results?term=
Refametinib&pg=1.
[
[
1] M. P. Dudukovic, F. Larachi, P. L. Mills, Chem. Eng. Sci. 1999, 54, 1975–
995.
2] a) R. Porta, M. Benaglia, A. Puglisi, Org. Process Res. Dev. 2016, 20, 2–25;
b) A. Adamo, R. L. Beingessner, M. Behnam, J. Chen, T. F. Jamison, K. F.
Jensen, J.-C. M. Monbaliu, A. S. Myerson, E. M. Revalor, D. R. Snead, T.
Stelzer, N. Weeranoppanant, S. Y. Wong, P. Zhang, Science 2016, 352, 61–
1
[10] P. Loos, H. Alex, J. Hassfeld, K. Lovis, J. Platzek, N. Steinfeldt, S. Hübner,
Org. Process Res. Dev. 2016, 20, 452–464.
67; c) D. M. Roberge, L. Ducry, N. Bieler, P. Cretton, B. Zimmermann,
Chem. Eng. Technol. 2005, 28, 318–323.
[11] M. Ben Said, T. Baramov, T. Herrmann, M. Gottfried, J. Hassfeld, S. Roggan,
[
3] a) T. Glasnov, Continuous flow chemistry in the research laboratory,
Springer, Cham, 2016; b) B. Gutmann, D. Cantillo, C. O. Kappe, Angew.
Chem. Int. Ed. 2015, 54, 6688–6728; Angew. Chem. 2015, 127, 6788; c) J.
Pesti, Org. Process Res. Dev. 2012, 16, 843–843; d) N. G. Anderson, Org.
Process Res. Dev. 2012, 16, 852–869; e) V. Hessel, Chem. Eng. Technol.
Org. Process Res. Dev. 2017, 21, 705–714.
[12] T. Baramov, P. Loos, J. Hassfeld, H. Alex, M. Beller, T. Stemmler, G. Meier,
M. Gottfried, S. Roggan, Adv. Synth. Catal. 2016, 358, 2903–2911.
[13] C. Perego, S. Peratello Foumeny, Catal. Today 1999, 52, 133–145.
[14] a) J. M. Devereux, K. R. Payne, E. R. A. Peeling, J. Chem. Soc. 1957, 2845;
b) H. Greenfield, J. Org. Chem. 1964, 29, 3082.
2009, 32, 1655–1681.
[
4] a) K. Geyer, T. Gustafsson, P. H. Seeberger, Synlett 2009, 2382–2391; b) C.
Wiles, P. Watts, Eur. J. Org. Chem. 2008, 1655–1671; c) T. Fukuyama, M. T.
Rahman, M. Sato, I. Ryu, Synlett 2008, 151–163; d) J. Yoshida, A. Nagaki,
T. Yamada, Chem. Eur. J. 2008, 14, 7450–7459; e) J. Wegner, S. Ceylan, A.
Kirschning, Chem. Commun. 2011, 47, 4583–4592.
[15] Q3D Elemental Impurities Guidance for Industry, http://www.fda.gov/
downloads/drugs/guidancecomplianceregulatoryinformation/guidances/
ucm371025.pdf, page 56.
[16] Supported by DIM-DAB still being detectable by HPLC even after pro-
longed washing of the catalyst bed with THF.
[
5] a) I. R. Baxendale, Org. Process Res. Dev. 2016, 20, 327–360; b) T. Razzaq,
C. O. Kappe, Chem. Asian J. 2010, 5, 1274–1289.
Received: May 4, 2017
Eur. J. Org. Chem. 2017, 3921–3928
www.eurjoc.org
3928
© 2017 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim