Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo synthesis enabled by online UHPLC analysis

Article abstract of DOI:10.1016/j.tet.2018.02.016

Process intensification commonly enables reaction acceleration and therefore continuous-flow/PAT is urgently needed. The low volumes typical for micro-flow pose challenges for online sampling operations in analytics. In this paper, a very fast process is combined with a modified ultra-high-performance liquid chromatography (UHPLC) system allowing for very fast sampling and analysis. Low-volume online sampling according to the needs posed by PAT for pharma quality control, is introduced here for UHPLC analysis of the photo-Claisen rearrangement in micro-flow. Chances and challenges are critically reviewed, including the reproducibility and robustness of the sampling. Furthermore, the ability and speed of the chosen set-up in order to capture process changes and adjust the process parameters properly is investigated. With the applied online sampling system, it was possible to perform, almost unattended and spending 12 times less sampling volume, a full factorial analysis of all relevant reaction conditions (243 experiments) in three days. Such quality-in-the-process-line (QuIProLi) online sampling avoided random errors due to automation.

Full text of DOI:10.1016/j.tet.2018.02.016

Tetrahedron xxx (2018) 1e9
Contents lists available at ScienceDirect
Tetrahedron
journal homepage: www.elsevier.com/locate/tet
Quality-In(Process)Line (QuIProLi) process intensification for a micro-
flow UV-photo synthesis enabled by online UHPLC analysis
a,
*,
,
, *
1, Elnaz Shahbazali a 1, Maarten Honing b, Volker Hessel a
ꢀ
Marc Escriba-Gelonch
a Eindhoven University of Technology, Department of Chemical Engineering and Chemistry, Group Micro Flow Chemistry and Process Technology, De
Rondom 70, 5612 AP Eindhoven, The Netherlands
b Maastricht University, Imaging Mass Spectrometry, Minderbroedersberg 4-6, 6211 LK Maastricht, The Netherlands
a r t i c l e i n f o
a b s t r a c t
Article history:
Process intensification commonly enables reaction acceleration and therefore continuous-flow/PAT is
urgently needed. The low volumes typical for micro-flow pose challenges for online sampling operations
in analytics. In this paper, a very fast process is combined with a modified ultra-high-performance liquid
chromatography (UHPLC) system allowing for very fast sampling and analysis. Low-volume online
sampling according to the needs posed by PAT for pharma quality control, is introduced here for UHPLC
analysis of the photo-Claisen rearrangement in micro-flow. Chances and challenges are critically
reviewed, including the reproducibility and robustness of the sampling. Furthermore, the ability and
speed of the chosen set-up in order to capture process changes and adjust the process parameters
properly is investigated. With the applied online sampling system, it was possible to perform, almost
unattended and spending 12 times less sampling volume, a full factorial analysis of all relevant reaction
conditions (243 experiments) in three days. Such quality-in-the-process-line (QuIProLi) online sampling
avoided random errors due to automation.
Received 7 November 2017
Received in revised form
25 January 2018
Accepted 5 February 2018
Available online xxx
Keywords:
Process automation
Flow chemistry
Fast-sampling
UHPLC
Photo-Claisen rearrangement
© 2018 Elsevier Ltd. All rights reserved.
1. Introduction
reducing labour and costs.9 In-time analytics are usually classified
depending on where and how the sampling operation is carried out
Over the last 20 years, continuous micro-flow reactions have
become an eminent topic in process design.1 The enhanced process
control given through miniaturization results in many benefits such
as higher yields, less waste, and improved sustainability.2 More far
fetching, flow chemistry makes meanwhile possible the develop-
ment of multi-step end-to-end platforms which deliver the product
from raw materials (up to the pill in pharmaceutical
manufacturing) in one continuous run. The need to keep control of
such cascaded processing has paved the way for finding a suited
analytical technology and fast detection which finally demand for
process automation,3 reducing the human operator interaction to
check a software interface.4e8 Nowadays the goal for multi-step
processes is to be fully telescoped and managed by a single
researcher.3 Therefore, continuous automated reaction monitoring
can enhance the reliability and the quality of the final product,
as: offline, atline, online and inline.10
Offline sampling in batch processing has some drawbacks even
when its single steps are as fast as continuous micro-flow. First, the
reaction outcome is often superimposed products by mass and heat
transfer and even dependent on the location of sampling in the
vessel, which is termed heterogeneity here. Therefore, in order to
get the complete view, a large sample volume has to be taken at
different points in the vessel. Complex statistical strategies are used
in quality control of industrial chemical production.11 In addition,
the heterogeneity of batch processes might cause higher errors or
higher deviations between samples, and therefore lower confi-
dence. These drawbacks are partially overcome sampling (offline)
in continuous flow due to the (spatial and timely) homogeneity of
sample qualities within the microcapillary. Thus a lower amount of
sampling and samples is needed. Nevertheless, offline sampling
still do not solve (i) the need for a minimum sampling volume to
allow the, also longer, manual work-up, (ii) the possibility to
damage or contaminate the sample during the operation, especially
with air-/light-sensitive or short lifetime compounds, and subse-
quently (iii) the chance of errors. Here, inline/online analysis
* Corresponding author.
E-mail addresses: m.escriba.gelonch@tue.nl (M. Escriba-Gelonch), v.hessel@tue.
ꢀ
nl (V. Hessel).
1
ꢀ
Dr. Marc Escriba Gelonch and Elnaz Shahbazali share the first authorship.
https://doi.org/10.1016/j.tet.2018.02.016
0040-4020/© 2018 Elsevier Ltd. All rights reserved.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
2
becomes a viable solution.
flow process monitoring.47 Other issues are the analysis costs,37 as
well as the long equilibration times when non-isocratic conditions
are used.40
Intensified continuous flow provides order-of-magnitude
shorter residence times12e14 and essentially needs ultrafast sam-
pling operating with lower volumes and without interruption. As a
consequence, the frequency of sampling and the time for analysis
simply has to be faster than the time needed for completion of
reaction and changing to the next process condition. Accordingly,
online and inline analysis is preferred over atline and offline, and
even those need to be speed up as given in this paper. An auto-
mation of online sampling also allows for the use of the precise
volume needed, avoiding the unprecise manual sampling, which
tends to increase the sampling volume in order to ensure enough
quantity. In addition, automation also reduces random errors
derived by manual operations, which take also more time to be
carried out.
Recognizing the ambition of the United States Food and Drug
(FDA) to establish continuous processing in pharmaceutical
manufacturing by 202615 and our previous flow chemistry
research,16e18 the application of Process Analytical Technology
(PAT)19,20 to flow chemistry is a major future goal relevant to the
innovation of the paper. Here, real-time information accessibility
has proven to enhance process control by reducing the timeline of
the measurements, with the possibility to correct any leverage or
malfunction in-time avoiding loses of chemicals. This creates a vast
amount of data which typically demands for some extra data pro-
cessing strategy. Established are multi-factorial mathematical
analysis, statistical evaluations, model predictions, and algorithms
like the Nelder-Mead simplex optimization algorithm.21 Common
to all is to aim to optimize the pattern recognition of the variables
or to improve even in two-dimensional fashion (e.g. temperature
and reaction time). The traditional one-factor-at-a-time sometimes
is not effective in explaining the interactions between factors.22,23
Coupling PAT systems to processes would avoid sample transfer
and therefore their possible deterioration. In order to allow
adequate monitoring and efficient reaction control, the sampling
time and subsequent analysis has to be very short compared to the
overall reaction time. In this connection, some analytical tech-
niques are preferred, such as near infrared (NIR),24e26 Raman,27
Mid-IR,28 acoustic emission signals,29 X-ray absorption spectros-
copy30 and nuclear magnetic resonance31,32 because of their
analytical speed, non-contact spectroscopic analysis as well as their
condition of not destructive methods. As an example, IR spectros-
copy is commonly used for the determination of physical factors
such as moisture,33 crystalline polymorphism,34 particle size,35 and
density.36
Nonetheless, the development of PAT for monitoring of micro-
flow chemistry has not largely reported. One reason might be the
small volumes of chemicals in the micro-channels which pose a
need for miniaturized versions of conventional analytical in-
struments.37 Therefore, process miniaturization requires an orders
of magnitude smaller volume sampling, which in micro process
engineering would mean easily most of the volume of the whole
reactor. Some analytical techniques, which include sampling, fulfil
that criterion such as Raman,38 liquid chromatography,39 or capil-
lary electrophoresis.40,41 However, for small molecule analysis,
high-performance liquid chromatography (HPLC) is commonly
preferred and is quasi the unique method used for process moni-
toring applications in micro-channels offline. HPLC coupled to
microreactors have been applied for the synthesis of, e.g., cyclo-
adducts,42 pyrazoles43 and ciprofloxacin analogues.44 Flow auto-
mation has also reached continuous multistep systems.45 The work
of Kock et al.46 shows one issue to be solved around automated flow
operation. 51 samples (0.2 mg solution) collection needed as much
as 5.6 h sampling time. Also, the length of the chromatographic
methods still is a main problem for the use HPLC for online micro-
Yet, with the already proven ultra-high-performance liquid
chromatography (UHPLC) conditions, analysis times are shortened
significantly. Together with micro column approaches, low sample
volumes are feasible, allowing fast sequential experiments, and also
novel process time windows to keep the quality-in-the-process-
line (QuIProLi). As an example, UHPLC allows to operate free of
constraints by heat transfer limitations, much different from batch
experiments.48 Indeed since recently there is evidence for
advanced process control with the goal of fast kinetic screen-
ing,49e53 biological analysis54e56 and mechanistic studies57,58 un-
der stable, well reproducible reaction conditions. The combination
of flow chemistry and online monitoring gives also the chance for
non-stop UHPLC continuous analytics, when spectroscopic online
analytic sources, e.g. FTIR or NMR, are not suitable because of (i)
specific products which overlap signals, (ii) reactions performed in
very low concentrations, or (iii) simply because in a microchannel
the amount of analyte is not enough for an accurate measurement.
As an example, online UHPLC coupled with a Teflon/FEP-capillary-
based flow microreactor has been reported as especially effective
for the analysis of non-volatile organic molecules because of the
rapid separation achieved.59 However and as outlined above, the
sampling step of UHPLC remains an issue, because of the small
volumes supplied by the small channels. Attiya et al.60 and Lin
et al.61 reported in 2001 the first approaches to online sampling
using micro-channels, but both applied to electrophoresis. Schlund
et al.62 reported in 2007 the first continuous sampling applied to
HPLC based on a double-T junction emulating an HPLC injector
loop. Also, fraction collectors have been used for sampling opera-
tion in microchannels.63 Therefore, the next step in micro-flow
based reaction development should be to determine the optimal
operating speed, as suggested by FDA.64
In this paper, the combination of flow chemistry with a modified
UHPLC system allowing rather fast online sampling is discussed. As
a proof-of-concept, this configuration was tested in combination
with a photo-Claisen rearrangement reaction. We previously re-
ported the thermal Claisen rearrangement of allyl phenyl ether
(APE) to ortho-allyl phenol (Fig. 1) achieving process intensifica-
tion65 without the need of a catalyst. Latter an intensification of
micro flow route for photo-Claisen rearrangement was also re-
ported.66 The photo pathway gave para-substituted isomer besides
of ortho-substituted given by thermal process. The latter opened
the chance of a new radicalary unexplored path. Therefore, the
photo-reaction path used in this study and given in Fig. 1, offers still
an attractive way for research in an automated fashion. Here, after
carrying out the reaction in the capillary photoreactor, samples are
automatically taken and subsequently analysed using short timing
enabled by the modified UHPLC system coupled to the setup. The
different residence times for the different operations, the number
of samples, the process simplification and the process automation
are parameters taken into account.
Fig. 1. Photo-Claisen rearrangement path used in online sampling.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
3
2. Experimental part
reactor. The sampling cycle was programmed to perform a system
purge, sampling and dilution followed by separation and detection.
2.1. Experimental setup
40
mL sample was trapped before being diluted with acetonitrile to
a total volume of 100
m
L. An aliquote of 1 L was subsequently
m
Microchannel experiments have been performed using the
reactor set-up depicted in Scheme 1. The setup is composed of two
HPLC pumps (Knauer Azura P4.1S). The first feeds a 0.5 M stock
solution of APE (Sigma Aldrich 99%,16.77 g,125 mmol) in 250 mL n-
butanol (Sigma Aldrich). The second feeds pure n-butanol and is
used to arrange, in-flow and in one run, the experiments carried out
using 0.1 M and 0.05 M concentration. A T-mixer (Valco) is subse-
quently used for this purpose. The flow-rate is adjusted accordingly
to achieve reaction times of 2, 5 and 8 min respectively. Inside an
oven, a low-pressure amalgam UV-lamp with distinctive peak at
254 nm (TS23-212; Dinies Technologies GmbH) is installed, and 5 m
(approx. 1 mL equivalent volume) UV-transparent fluorinated
ethylene propylene capillary tubing (FEP 1548, 0.5 mm ID;
Upchurch Scientific) is wound on a metallic tower around the cy-
lindrical shaped lamp. Accordingly, the tubing is properly con-
nected to the T-mixer in the inlet and to the sampling unit in the
outlet of the oven. Just before online UHPLC connection, a T-piece
(IDEX) allows for offline sampling.
injected onto the analytical column.
3. Results and discussion
3.1. Online sampling and analytic time distribution
The first goal was to investigate three experimental parameters
with certain discrete values for the photo-Claisen reaction as fast as
possible using the online UHPLC-UV analyzer, i.e. to show minimal
analysis time demands and to benchmark this to any combination
of conventional and micro-flow processing with conventional an-
alytics. These three parameters were considered: temperature (T),
concentration (c) and reaction time (t). Concerning temperature,
the values were 21 ꢁC, 35 ꢁC and 50 ꢁC, for concentration 0.05 M,
0.1 M and 0.5 M, and for reaction times were 2 min, 5 min and
8 min. Each operation was timed using a chronometer.
3.1.1. Timing of sampling, analysis, and settling time for next
experiment
2.2. Experimental conditions of sampling and chromatographic
separation
Fig. 2 gives the time distribution for testing three flow rates at
one temperature and one concentration, here referred as one cycle.
In this Figure, the fragment striped refers to the settling and sta-
bilization time (here, three residence times were taken). Also the
dead time refers to the delay of setting the conditions in flow.
Despite of the advantages of flow chemistry in terms of heat and
mass transfer, the existence of such dead time frames suggest that
flow chemistry could not be fast enough in this case. After setting
them three online samples were taken, all of them with the same
protocol. In total, the sampling operation (including sampling and
automatic work-up) lasted 7 min. After sampling, a new flow rate
was set. During the stabilization time, the UHPLC analysed auto-
matically all samples and the results were taken before the new
sampling (Fig. 3). In addition, during the next sampling operation,
the results of the previous samples were processed by the user
(Fig. 3). As a consequence, 9 samples in 3 different conditions were
taken, analysed in parallel and the results were processed. The
online test speed then was set at 72 min/cycle. It is relevant to note
that the time for sampling and analysis was usually shorter than the
time for setting new conditions. Then this ultrafast procedure
Allyl phenyl ether, 1-butanol, formic acid, acetonitrile, and 1,3-
dinitrobenzene were purchased from Aldrich chemical company
and used as received. The solvents purity were all according to
HPLC grade of Sigma-Aldrich solvents (>99.5). The online UHPLC
system was based on an Agilent 1290 Infinity II LC system (Scheme
1) consisting of a binary pump, an autosampler, a thermostated
column compartment and a diode array detector equipped with a
10 mm Max-Light Cartridge Cell. Analytical isocratic separation was
achieved with an Agilent ZORBAX Eclipse Plus C18 UHPLC column
(2.1 ꢀ 50 mm, 1.8
mm) and applying a mixture of 0.1% formic acid in
water and acetonitrile (55/45; v/v) as the mobile phase. The
detection was carried out with a UVevisible diode array detector at
270 nm, and data processing performed with the Agilent OpenLAB
ChemStation software. The analytical setup was linked to the flow
reactor by means of an “in-house” modified autosampler (an item
of a patent application), encompassing a series of multi-port valves,
which allows the direct sampling from the outlet of the flow
Scheme 1. Photo-Claisen rearrangement setup coupled to offline and online sampling. 1: Stock solution of alkyl phenyl ether in n-butanol. 2: n-butanol. 3: online autosampling. 4:
Automated dilution. 5: UHPLC analyzer.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
4
Fig. 2. Timeline for sequentially testing 3 flow rates at the same temperature and concentration. Online sampling allowed for, 9 samples (taken and analysed) in 72 min. Offline
sampling needed around 30 min more. The fragment striped refers to the time needed to set and stabilize the new conditions (3 residence times).
Fig. 3. Online sample chromatogram of 0.05 M APE, using 2 min reaction time at 21 ꢁC.
becomes faster than continuous flow intensification. In addition,
the researcher only needed to set the conditions and get the results,
which reduced drastically the random error derived of manual
operation. Thus, three flow rates were triple tested largely auto-
mated in somewhat more than one hour ten minutes. In this way,
for each of the 3 flow rates, 3 temperatures and 3 concentrations (9
cycles) were tested in just a bit more than 10 h in the same day.
Thus, the parametric variations sum up to 3 ꢀ 3 x 3 ¼ 27 mea-
surements. As all of them were repeated three times, 81 samples
were achieved in one run and one day. The same set of experiments
was carried out in 3 different days, during which 27 conditions
were three times tested and 243 samples were taken, analysed and
the results were processed.
3.1.2. Sampling volume
The same set of experiments was carried out offline. Here both
sampling and dilutions were performed manually. The sampling
time was set according to the reactor flow-rate in order to achieve
0.5 mL sample. An important point to highlight is the fact that this
sampling volume means almost the half of the volume of the whole
capillary microreactor. In addition, online sampling needed 12
times less volume to perform the same experiments, as shown in
Fig. 4, which is more suitable for low volumes used in flow chem-
istry. Batch sampling in Fig. 4 refers to 5 sampling points in a batch
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
5
online sampling delivered 4.5 h less, therefore in overall, online
experiments saved 40% time.
3.1.3. Sampling reliability and robustness
Table 1 shows the average values, their standard deviation and
their variation coefficient (VC) of each set of experiments per-
formed in the 3 days. The repeatability is characterized by the VC
and found in all cases to be below 3% using low and medium
concentrations. Nevertheless, when increasing the concentration,
the variability was found to be higher, especially in low residence
times. This finding could be explained by (i) the viscosity of the final
solution at high concentration (viscosity of just n-butanol is rather
high: 2.6 mPa s at 25 ꢁC),68 which can modify the rheology and the
hydrodynamics making difficult to achieve a homogeneous sam-
pling at the same drawing speed. Furthermore, (ii) the precision
could be compromised because of solubility constraints due to the
incomplete reaction of precipitated ortho allylphenol. In fact, it was
observed that at high concentrations the solution sometimes
became not fully homogeneous because of the close solubility limit.
This could be a limitation for representative sampling, since an
online particulate matter of 40 mL might be not sufficient compared
to 0.5 mL in offline sampling. The viscosity- and solubility-related
explanation is in line with the finding that the repeatability is
improved with an increase of temperature.
Table 2 summarizes the reproducibility of these results. The
term reproducibility is used, as the same experiment carried out
under the same conditions but at different days. Then, Table 2 gives
the deviation between the mean values of the same three repro-
ducible experiments carried out in different days. The majority of
the VC values are below 10%, except the ones at higher concen-
tration. Rheological and solubility issues may explain these results
as described above.
Fig. 4. Comparison between the needed volumes for every set of 3 samples.
reactor (4 h irradiation time according to Pitchumani et al.67). In
continuous flow, for 2, 5 and 8 min residence time, 1 min/sample
3 min/sample and 4.25 min/sample were needed respectively.
Therefore, in Fig. 5 it is noted that the difference between online
and offline sampling increases when the flow rate in the reactor
decreases, because of the time needed offline to get 0.5 mL of
sample even with low flow rates. The subsequent work-up of three
samples in order to get the final injection vials was set in 6 min. This
operation is reduced drastically online in terms of time, avoiding
also random errors. Overall, offline sampling operation and work-
up needed 10 min, 16 min and 20 min for 2, 5 and 8 min reaction
time respectively for each sample. Fig. 2 shows the compared
timeline online/offline for each cycle of experiments. The difference
in time was set to almost half an hour for each cycle, which means
3.2. Analysis of the confidence in the results obtained
As described above, simultaneously to the online sampling,
conventional samples in 5 mL vials (VWR International) were taken
in parallel for each set of conditions. These samples were manually
diluted and injected in the same UHPLC using the conventional
procedure of injection from vials. These results were used to
compare offline sampling and sample preparation with online
sampling. Hence, each online average and deviation value was
opposed to the corresponding offline average and deviation value.
The subtraction of these paired results generated a new variable
defined as W ¼ online e offline. Statistical procedures were used to
Fig. 5. Comparison between sampling procedures for a single set of 3 samples at low, medium and high residence times, equivalent to high, medium and low flow rates respectively.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
6
Table 1
Average values of three repeated online measurements of photo-Claisen reaction conversions obtained at three different days. (VC ¼ variation coefficient).
Concentration (M) /
0,05
0,1
0,5
Day
I
T (ꢁC)
21
t (min)
Conversion (%)
Std. Dev.
% VC
Conversion (%)
Std. Dev.
% VC
Conversion (%)
Std. Dev.
% VC
2
5
8
2
5
8
2
5
8
2
5
8
2
5
8
2
5
8
2
5
8
2
5
8
2
5
8
49.7
82.5
94.5
51.0
83.4
94.6
59.9
89.2
94.0
42.5
72.5
87.6
43.0
72.6
87.8
65.0
91.0
95.5
46.1
77.9
91.5
45.0
76.8
90.9
58.0
86.7
97.5
0.772
0.228
0.047
0.409
0.167
0.027
0.611
0.422
0.203
0.303
0.710
0.190
0.463
0.369
0.270
0.027
0.044
0.016
1.454
0.377
0.038
0.417
0.037
0.162
0.118
0.187
0.103
2
0
0
1
0
0
1
0
0
1
1
0
1
1
0
0
0
0
3
0
0
1
0
0
0
0
0
30.4
55.5
72.4
31.9
57.0
74.4
39.0
71.0
85.9
31.0
50.5
69.0
30.0
55.0
70.6
38.6
68.1
83.0
33.2
56.0
74.0
29.8
56.5
73.1
38.1
68.0
85.5
0.916
0.563
0.321
0.102
0.025
0.096
0.025
0.015
0.015
0.040
0.305
0.238
0.070
0.026
0.544
0.717
0.040
0.053
0.614
0.093
0.091
0.817
0.684
0.091
0.220
0.218
0.235
3
1
0
0
0
0
0
0
0
0
1
0
0
0
1
2
0
0
2
0
0
3
1
0
1
0
0
6.0
1.565
1.267
0.227
1.561
1.361
1.118
0.428
0.071
0.041
1.042
0.335
0.697
1.811
1.250
0.673
0.463
0.250
0.034
1.399
0.962
1.171
1.797
0.933
0.054
0.303
0.105
0.036
26
8
1
18
7
4
4
0
0
14
2
4
36
9
3
5
1
0
22
6
5
15.7
23.1
8.8
20.0
27.6
9.5
22.0
32.0
7.3
13.8
19.3
5.1
13.8
23.0
10.0
19.9
30.0
6.3
17.1
23.6
7.2
18.7
23.9
11.0
22.0
33.5
35
50
21
35
50
21
35
50
II
III
25
5
0
3
0
0
Table 2
Averages of three repeated online measurements of photo-Claisen reaction conversions obtained at different days.
Concentration (M)
/
0.05
0.1
0.5
T (⁰C)
t (min)
Conversion (%)
Std. Dev.
%VC
Conversion (%)
Std. Dev.
%VC
Conversion (%)
Std. Dev.
%VC
21
2
5
8
2
5
8
2
5
8
46.1
77.6
91.2
46.4
77.6
91.1
60.9
89.0
95.6
3.602
5.006
3.480
4.166
5.479
3.376
3.618
2.136
1.741
8
6
4
9
7
4
6
2
2
31.5
54.0
71.8
30.6
56.1
72.7
38.6
69.0
84.8
1.477
3.070
2.577
1.178
1.032
1.880
0.454
1.718
1.568
5
6
4
4
2
3
1
2
2
6.6
0.690
1.633
2.388
1.868
3.258
2.441
0.752
1.209
1.743
10
10
11
26
19
10
7
15.6
22.0
7.0
17.5
24.8
10.2
21.3
31.8
35
50
6
5
study this new variable using Statgraphics Centurion VII. In ideal
conditions, the average of this variable should be 0 with 0 devia-
tion, which means that the result obtained by online would be
exactly the same, as the corresponding offline sample. Fig. 6 shows
the frequency histogram of W, where a Gauss bell shows a slight
tendency of a leverage in the 0 value of W. Fig. 7 shows the normal
test which suggests that the values of W can be represented by a
normal distribution at 95% of confidence level. The latter is relevant
in order to carry out further hypothesis tests and set the interval of
confidence of W.
In a context of normality, a hypothesis test was performed in
order to determine if it could statistically be accepted that the
average of W is zero, and therefore the results online are as reliable
as offline. The null hypothesis of the T-test was subsequently:
average equal to 0, giving an alternative hypothesis of non-equality.
Unfortunately, this test was rejected with 95% confidence level,
concluding that the average of W is not 0, and therefore a sys-
tematic leverage has to be accepted. In order to delimit this
leverage, a new test was performed based on the distribution of W
showed in Fig. 6. The assumption that the average value equals 0.5
Fig. 6. Frequency histogram of variable W.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
7
significance of the model and Table 3 (down) shows the analysis of
variance (ANOVA) derived from the statistical analysis. Here, the F-
statistic refers to a ratio of two variances: variation between sample
means and variation within the samples. Thus, higher dispersion
gives higher F value, which means that concentration gives more
variability on the APE conversion than reaction time and
temperature.
Table 3 shows that all single variables: T, c and t have significant
(P < 0.001) differences in their levels. Therefore, the APE conversion
of the photo-Claisen reaction is different at different concentra-
tions, different temperatures and different residence times.
Nevertheless, apparently the interaction T * t is not significant,
which means that the variability of the conversion when using
different residence times is statistically the same when using
different temperatures. Hence, parallel lines are expected in the T *
t interaction plot (Fig. 8). In this plot, the 21 ꢁC and 35 ꢁC lines are
close to each other, which can be correlated with a statistically
identic behaviour related to the conversion. The latter is confirmed
by the Duncan's test (Table 4), which means that the same con-
version profile is obtained by operating at 21 ꢁC and at 35 ꢁC. In the
light of these results, it can be concluded that maximal APE con-
version of the photo-Claisen reaction is achieved at high temper-
ature (50 ꢁC), low concentration (0.05 M) and long residence time
(8 min). This is also shown on the 3D plot in Fig. 9.
Fig. 7. Normality test for W.
was checked. As a result, the P-value of this test was 0.0677836
which is above the significance level of 0.05. Hence, in these con-
ditions the null hypothesis cannot be rejected at the 95% confidence
level. Therefore, it can be accepted that the average value of W
is þ0.5, which means that the systematic difference for the online
sampling compared to manual sampling and dilution is þ0.5%.
In order to set the interval of confidence of W, a new hypothesis
test was performed to fix the standard deviation. In this case a Chi-
squared test69 was performed. Considering Fig. 6, the null hy-
pothesis was set to be equal to 1.4 versus the alternative hypothesis
of different. This test gave a P-value of 0.138879, above the signif-
icance level of 0.05. Hence the value of 1.4 can be accepted as the
standard deviation of W at the 95% confidence level. Therefore, it
can be concluded the variable W is normally distributed and has an
average of 0.5 with a standard deviation of 1.4. Transferring this
conclusion to the sampling, the online sampling gives 0.5% higher
conversion 1.4% compared to offline sampling.
4. Conclusions
In this paper, a novel analytical approach in online sampling,
separation and detection (QuIProLi) successfully applied for the
monitoring of a photo induced Claisen rearrangement reaction in
microflow channels is described. A single analysis/sampling into a
process-systemic approach is considered, considering also statis-
tical approaches and needs posed by PAT for pharma quality con-
trol. As the FDA has decided to push the pharmaceutical companies
to the transition from batch to flow, the establishment of a
continuous-flow/PAT is urgently needed. The new sampling unit
allowed the low volume and high speed sampling while main-
taining very good repeatability and reproducibility of both the
separation and detection of all reaction products. The developed
analytical set-up avoids the complex sampling strategies of batch
operation, as well as reduces drastically the manual work-up timing
as well as the sampling volume. The increased sampling speed,
allowed the screening of up to 27 process conditions in 24 h; due to
3 times repetition, 81 samples were taken. In average, one sampling
(at coupled continuous micro-flow reaction) took less than 7 min.
This fact becomes relevant compared to offline sampling especially
in low flow rates. Testing reproducibility via comparing tests at 3
3.3. Statistical approach of the Photo-Claisen reaction parameters
Taking advantage of the experiments carried out, an additional
study of the effect of the reaction conditions in the conversion of
the photo-Claisen reaction was performed, using one of the sets of
online experiments which corresponds to almost one day working.
In this connection, a balanced completely randomized design was
performed using the 3 levels of each factor (T, c and t) as described
above. All data were analysed using generalized linear model (PROC
GLM) of SAS (V9.0, SAS Institute Inc., Cary, NC, USA).70 The signifi-
cance of the analysis of variance was set at 0.05 and a Duncan's
multiple comparison test was added.71 Table 3 (up) shows the
Table 3
Significance of the GLM model (up) and the ANOVA derived of this analysis (down).
Source
DF
Sum of Squares
Mean Square
F Value
360.28
Pr > F
Model
Error
Corrected Total
26
54
80
66792.62840
385.04667
67177.67506
2568.94725
7.13049
<.0001
Source
DF
Type I/III SS
Mean Square
F Value
Pr > F
T
C
2
2
4
2
4
4
8
1477.26247
46782.28469
120.27457
16290.81654
14.45383
738.63123
23391.14235
30.06864
8145.40827
3.61346
103.59
3280.44
4.22
1142.33
0.51
<.0001
<.0001
0.0048
<.0001
0.7309
<.0001
0.0054
T * C
t
T * t
C * t
T * c * t
1928.20049
179.33580
482.05012
22.41698
67.60
3.14
Fig. 8. Interaction plot of residence time (t) and temperature (T) using 0.05 M
concentration.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
8
Table 4
Duncan's test for the temperature, concentration and residence time (means with the same letter are not significantly different).
Grouping
Conv. (%)
N
T (⁰C)
Grouping
Conv. (%)
N
c (M)
Grouping
Conv. (%)
N
t (min)
A
B
B
55.7074
47.0852
46.2667
27
27
27
50
35
21
A
B
C
75.0630
56.5778
17.4185
27
27
27
0.05
0.1
0.5
A
B
C
65.1000
53.0926
30.8667
27
27
27
8
5
2
8. Pfeiffer SA, Nagl S. Methods Appl Fluoresc. 2015;3:34003.
9. Carlson R, Carlson JE. In Design and Optimization in Organic Synthesis. second ed.
New York: Elsevier; 2005.
10. Minnich C, Hardy S, Kraemer S. Chem Ing Tech. 2016;88:694.
11. Process or product monitoring and control. In: NIST/SEMATECH e-handbook of
statistical methods. NIST; 2012.
12. Hessel V, Vural-Gursel I, Wang Q, Noel T, Lang J. Chem Eng Technol. 2012;35:
1184.
13. Hessel V. Chem Eng Tech. 2009;32:1655.
€
14. Hessel V, Kralisch D, Kockmann N, Noel T, Wang Q. ChemSusChem. 2013;6:746.
15. FDA. Pharmaceutical CGMPs for the 21st Centuryda Risk-Based Approach. Final
Rep. Rockv. MD. 2004.
16. Escriba M, Hessel V, Rothstock S, et al. Green Chem. 2011;13:1799.
17. Borukhova S, Noel T, Metten B, de Vos E, Hessel V. Green Chem. 2016;18:4947.
18. Borukhova S, Noel T, Hessel V. ChemSusChem. 2016;9:67.
19. Database of ICH Guidelines, Compilation prepared by International Conference
on Harmonisation of Technical Requirements for Registration of Pharmaceu-
ticals for Human Use (ICH), http://www.ich.org/products/guidelines (last ac-
cess 29 September 2017).
20. Database of FDA Guidance for Industry, Compilation prepared by Food and
Drug Administration, https://www.fda.gov/RegulatoryInformation/Guidances/
ucm081818.htm (last access 29 September 2017).
21. McMullen JP, Jensen KF. An intelligent microreactor system for real-time
optimization of a chemical reaction. In: Proc. Int. Conf. Miniat. Syst. Chem. Life
Sci. (mTAS), 12th; Chem. Biol. Microsyst. Soc.: San Diego. 2008:1908.
22. Montgomery DC. In Design and Analysis of Experiments. New York: John Wiley;
2001.
Fig. 9. Effect on conversion of the interaction of residence time (t) and Temperature (T)
using 0.05 M concentration.
23. Huyer W, Neumaier A. ACM Trans Math Softw. 2008;35:1.
24. Tanaka R, Takahashi N, Nakamura Y, et al. Anal Sci. 2017;33:41.
25. Roggo Y, Chalus P, Maurer L, Martinez CL, et al. J Pharm Biomed Anal. 2007;44:
683.
different days, all in all 243 samples were collected online, analysed
and processed under almost automatized conditions. The results
were enabled due to the speed of both online sampling and online
UHPLC analysis. As the latter is established, it can be concluded that
very fast online sampling opens the door for further intensification
analysis for flow chemistry. Here the automatic online sampling
was 40% faster than offline sampling, keeping the same repeat-
ability and robustness, and using more than 12 times less sample
volume than used offline. Therefore, in the studied case, assuming a
known systematic difference of 0.5% 1.4 compared to offline
sampling and analysis, online analysis can be taken with the same
reliability, avoiding the difficult to balance random errors. In
addition, the photo-Claisen reaction conditions in terms of con-
version were monitored and successfully optimized at 50 ꢁC, 0.05 M
and 8 min with the results obtained in just three working days.
26. Ferstl W, Klahn T, Schweikert W, et al. Chem Eng Technol. 2007;30:370.
27. Hausman DS, Cambron RT, Sakr A. Int J Pharm. 2005;298:80.
28. Floyd TM, Schmidt MA, Jensen KF. Ind Eng Chem Res. 2004;44:2351.
29. Hu Y, Wang L, Huang X, Qian X, et al. IEEE Trans Instrum Meas. 2015;64:1100.
30. Hübner S, Bentrup U, Budde U, et al. Org Process Res Dev. 2009;13:952.
31. Araneda JF, Boehringer T, Rehm T, Riegel S. 254th ACS National Meeting &
Exposition, Washington, DC, Aug 20e24. 2017.
32. Maguire Y, Chuang IL, Zhang S, Gershenfeld N. Proc Natl Acad Sci. 2007;104:
9198.
33. Otsuka M, Koyama A, Hattori Y. RSC Adv. 2014;4:17461.
34. Kamada K, Yoshimura S, Murata M, et al. Int J Pharm. 2009;368:103.
35. Tanabe H, Otsuka K, Otsuka M. Anal Sci. 2007;23:857.
36. Otsuka M. Powder Technol. 2004;141:244.
37. Koch MV, Marquardt B, Bonne U, et al. Selected Developments in Micro-
analytical Technology. In Micro Instrumentation: for High Throughput Exper-
imentation and Process Intensification - a Tool for PAT; Koch, MV, Vanden-
bussche, KM, Chrisman, RW, Ed; Wiley-VCH Verlag GmbH
& Co. KGaA:
Weinheim, Germany. https://doi.org/10.1002/9783527610624.ch9.
38. Marquardt B. Application of on-line Raman spectroscopy to characterize and
optimize a continuous microreactor. In: Koch MV, ed. Micro Instrumentation.
Weinheim, Germany: Wiley-VCH Verlag GmbH & Co. KGaA; 2007.
39. Malherbe CJ, de Beer D, Joubert E. Int J Mol Sci. 2012;13:3101.
40. McMullen JP, Jensen KF. Annu Rev Anal Chem. 2010;3:19.
41. Albrecht JW, El-Ali J, Jensen KF. Anal Chem. 2007;79:9364.
42. Fernandez-Suarez M, Wong SYF, Warrington BH. Lab Chip. 2002;2:170.
43. Garcia-Egido E, Spikmans V, Wong SYF, Warrington BH. Lab Chip. 2003;3:73.
44. Schwalbe T, Kadzimirsz D, Jas G. QSAR Comb Sci. 2005;24:758.
45. Griffiths-Jones CM, Hopkin MD, Jonsson D, Ley SV, et al. J Comb Chem. 2007;9:
422.
46. Koch K, van Weerdenburg BJA, Verkade JMM, Nieuwland PJ, Rutjes FPJT, van
Hest JCM. Org Process Res Dev. 2009;13:1003.
47. Shih CY, Chen Y, Xie J, et al. J Chromatogr A. 2006;1111:272.
48. McMullen JP, Jensen KF. Org Process Res Dev. 2010;14:1169.
49. Haven JJ, Zaquen N, Rubens M, Junkers T. Macromol React Eng. 2017;11,
1700016.
Acknowledgements
The authors acknowledge Ms. Bettina Schuhn, Mr. Thomas
Ortmann and Dr. Thomas Glauner (Agilent Technologies) for sup-
porting the modification of the UHPLC system for online sampling.
This project was funded by the Horizon 2020: Marie Skolodowska-
ꢀ
Curie Individual Fellowship awarded by Dr. Marc Escriba Gelonch
under grant agreement number 659233.
References
50. Murphy ER, Martinelli JR, Zaborenko N, Buchwald SL, Jensen KF. Angew Chem
Int Ed. 2007;46:1734.
1. Plutschack MB, Pieber B, Gilmore K, Seeberger PH. Chem Rev. 2017;117:11796.
2. Mason BP, Price KE, Steinbacher JL, et al. Chem Rev. 2007;107:2300.
3. Ley SV, Fitzpatrick DE, Ingham RJ, Myers RM. Angew Chem Int Ed. 2015;54:3449.
4. Sans V, Cronin L. Chem Soc Rev. 2016;45:2032.
5. Reizman BJ, Wang YM, Buchwald SL, Jensen KF. React Chem Eng. 2016;1:658.
6. McMullen JP, Stone MT, Buchwald SL, Jensen KF. Angew Chem Int Ed Engl.
2010;49:7076.
51. Ratner DM, Murphy ER, Jhunjhunwala M, et al. Chem Commun. 2005;5:578.
52. Tadepalli S, Halder R, Lawal A. Chem Eng Sci. 2007;62:2663.
53. a) Floyd TM, Schmidt MA, Jensen KF. Ind Eng Chem Res. 2004;44:2351;
b) Quiram DJ, Jensen KF, Schmidtt MA, et al. Ind Eng Chem Res. 2007;46:8319.
54. El-Ali J, Sorger PK, Jensen KF. Nature. 2006;442:403.
55. Gerber D, Maerkl SJ, Quake SR. Nat Methods. 2009;6:71.
7. Sans V, Porwol L, Dragone V, Cronin L. Chem Sci. 2015;6:1258.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016

ꢀ
M. Escriba-Gelonch et al. / Tetrahedron xxx (2018) 1e9
9
56. Srisa-Art M, Kang DK, Hong J, Park H, et al. ChemBioChem. 2009;10:1605.
64. Yu L. Pharm Res. 2008;25:781.
57. Ajmera SK, Delattre C, Schmidt MA, Jensen KF. J Catal. 2002;209:401.
58. Floyd TM, Schmidt MA, Jensen KF. Ind Eng Chem Res. 2004;44:2351.
59. Fang H, Xiao Q, Wu F, et al. J Org Chem. 2010;75:5619.
60. Attiya S, Jemere AB, Tang T, Fitzpatrick G, et al. Electrophoresis. 2001;22:318.
61. Lin YH, Lee GB, Li CW, Huang GR, Chen SH. J Chromatogr A. 2001;937:115.
62. Schlund M, Gilbert SE, Schnydrig S, Renaud P. Sens Actuators B Chem. 2007;123:
1133.
65. Shahbazali E, Spapens M, Kobayashi H, et al. Chem Eng J. 2015;281:144.
66. Shahbazali E, Noel T, Hessel V. J Flow Chem. 2016;6:252.
67. Pitchumani K, Warrier M, Ramamurthy V. J Am Chem Soc. 1996;118:9428.
68. n-Butanol. BASF petrochemicals (M 2084 e) edition; July 2014. http://www.
solvents.basf.com/portal/streamer?fid¼278912. Accessed October , 2017.
69. Pearson K. Philos Mag Series. 1900;5:50e157.
70. SAS institute. Base SAS 9.2 Procedures Guide, Statistical Procedures. second ed.
Cary, NC: SAS Institute; 2009.
€
63. Koch K, van Weerdenburg BJA, Verkade JMM, Nieuwland PJ, et al. Org Process
Res Dev. 2009;13:1003.
71. Duncan DB. Biometrics. 1955;11:1.
ꢀ
Please cite this article in press as: Escriba-Gelonch M, et al., Quality-In(Process)Line (QuIProLi) process intensification for a micro-flow UV-photo
synthesis enabled by online UHPLC analysis, Tetrahedron (2018), https://doi.org/10.1016/j.tet.2018.02.016