PROCESS DEVELOPMENT AND CASE STUDIES OF CONTINUOUS REACTOR SYSTEMS FOR PRODUCTION OF API 333

14.6.2 Process Analytical Technology (PAT) approximately 0 C. The complete reaction mixture was
immediately quenched into a biphasic mixture of aqueous
Process analytical technology (PAT) is an important part of
base and ethyl acetate. An amide impurity was formed at high
most continuous processes as it provides a useful means of
levels of greater than 2%. The longer quench times antici-
monitoring the state of the reaction. Indeed, one of the stated
pated upon scale-up were expected to further increase the
goals of the FDA’s PAT initiative is “Facilitating continuous
level of the amide impurity.
processing to improve efficiency and manage variability”
A continuous processing approach was undertaken to
[23]. Typical PAT tools include Raman, FTIR, NIR, and
minimize impurity formation through improved control of
UV–Vis spectroscopy [24, 25] or other noninvasive monitor-
reaction time and reduced quenching time. The continuous
ing techniques that can be adapted to a flow cell or tube reac-
reaction was assessed in the laboratory by mixing two feed
tor. These tools can be used during both transient and steady-
streams, one for the substrate and the other TFA, in a glass
state operations. As an example, FTIR was used to determine
microreactor with an overall volume less than 10 ml. Experi-
the proper ratio of reagent feed rate to starting material feed
ments varying temperature and residence time identified
rate during the startup of a continuous process to make an
process conditions, 25 C and a minimum residence time
active pharmaceutical ingredient [18]. For this particular
of 4 minutes, which provided complete conversion and a sig-
process, the same PAT equipment could have been used to
nificantly lower level of amide impurity, approximately 1%.
monitor the product quality. In a well-defined continuous
The preliminary reaction kinetics obtained from the small-
process, it is envisioned that feedback controllers could
scale continuous reactions paved the way for a rapid process
adjust operating parameters based on input signals from
scale-up. A 100-fold increase in flow rate in the substrate and
PAT analyzers. Even in the absence of feedback control,
TFA streams was used to process approximately 5 kg of start-
PAT can provide valuable information to plant operators
ing material using the setup shown in Figure 14.11. The start-
who can modify the operation if necessary. If PAT analyzers
ing material solution was not stable at room temperature and
indicate that product quality is suspect, flow can be diverted
required storage at −10 C. A preconditioning heat exchanger
to alternative holding tanks for further analysis. In the
was used to continuously heat up the starting material feed
absence of spectroscopic analyzers, simple temperature mea-
stream to the reaction temperature, 25 C, just prior to reac-
surements at various reactor positions can provide a wealth of
tion. A PFR, constructed of a jacketed static mixer for mixing
information regarding reaction performance.
the two feeds and three sequential concentric tube heat
exchangers, operated with an overall residence time of five
minutes. The reaction stream was continuously quenched in a
14.7 CASE STUDY: CONTINUOUS jacketed static mixer, and the quenched mixture flowed into a
DEPROTECTION REACTION – LAB TO KILO receiver for subsequent processing. The use of a static mixer
LAB SCALE-UP for the quench ensured effective mixing of the biphasic proc-
ess stream while rapidly quenching the reactive species.
A batch process to carry out an acidolysis and deprotection This particular batch process was relatively simple to con-
chemistry for a pharmaceutical intermediate involved adding vert to a continuous process. However, it is a good example to
the substrate solution to triflouroacetic acid (TFA) at demonstrate the key components and strategies behind the

Starting material Processing rate: 1.1 kg starting material/hour

–10 °C

Plug flow reactor Static

(~5 minutes residence time) mixer

–10 °C
50 ml/min
25 °C HX
25 °C

TFA Static
mixer Quench/
base

Receiving tank
70 ml/min with EtOAc
276 ml/min

FIGURE 14.11 Kilo lab continuous flow setup for acidolysis and deprotection process.
334 CHEMICAL ENGINEERING IN THE PHARMACEUTICAL INDUSTRY

development process. This includes the use of stable feeds, degradation. Initial screening work utilized a coiled 1/8 inch
preconditioning of a feed, and combined in-line jacketed stainless steel jacketed tube as the reactor. Later in development,
static mixer and heat exchangers as the reactor. multiple 26 ml shell and tube heat exchangers containing up to
19 1/8 inch stainless steel tubes (Figure 14.12) were utilized.
The heat exchangers were sequenced end to end to form the
14.8 CASE STUDY: CONTINUOUS PFR. The reactor was operated with a short 50 second residence
PRODUCTION OF A CYCLOPROPONATING time. Due to the short residence time and the large amount of
REAGENT heat generated early in the reactor, isothermal operation was
not possible. Details of the process are described below.
14.8.1 Introduction The laboratory setup used for development of the contin-
The Simmons–Smith cyclopropanation is a well-known reac- uous process is shown in Figure 14.13. Two feed streams,
tion to form cyclopropanes from olefins utilizing zinc and an one containing diethyl zinc (13.5 wt %) and dimethoxyethane
alkyl iodide. The structure of the reactive zinc carbenoid spe- in toluene and the other containing diiodomethane in dichlor-
cies is the subject of numerous papers [26, 27]. Formation of omethane, were held at ambient temperature. These streams
the active species is relatively exothermic with an adiabatic were fed to the reactor with gear pumps and mass flow meters
temperature rise above 120 C. Additionally, the complexes to ensure proper stoichiometry and residence time. Both
are known to be unstable for extended periods of time above streams passed through independent heat exchangers with
0 C. The exothermic nature of the reaction, combined with a 30 C jacket temperature prior to mixing in a non-jacketed
the complexity and incomplete understanding of the mechan- static mixer containing 27 helical mixing elements. The feeds
ism, and relative instability of the active species made scale- entered the static mixer at about 29 C and exited at about
up very challenging in a batch process. One solution to the 48 C. The reaction mixture flowed through a series of three
scale-up was the development of a continuous process for shell and tube heat exchangers with 30 C jacket temperature
formation of the cyclopropanating reagent. The process to facilitate formation of the reagent. The process stream exit
was demonstrated at laboratory scale, scaled up to pilot plant temperature was 52 C after the first heat exchanger and 32
scale, and was used to make launch supplies for the starting C after the third. The 2 C temperature difference observed
material of a commercial API. The development and imple- between process and jacket sides of the third heat exchanger
mentation of this process are discussed here. indicates that the reaction was nearly complete by that stage.
Additional calorimetric laboratory experiments that evalu-
ated residual heat generation of the reaction mixture con-
14.8.2 Process Development
firmed this observation. Finally, the reaction was thermally
The strategy for developing a continuous process was to operate quenched to less than −10 C, again with a shell and tube heat
a PFR with a short residence time and higher temperatures, fol- exchanger. The process attained a steady state within three
lowed by a rapid thermal quench. A short reaction time was residence times based on multiple temperature measurements
required to minimize the size of the PFR as well as reagent at various points along the PFR.

163 ×1/8 inch tubes

26 ml lab unit

420 ml pilot plant unit

FIGURE 14.12 Mini shell and tube heat exchangers for laboratory or pilot plant use.
 PROCESS DEVELOPMENT AND CASE STUDIES OF CONTINUOUS REACTOR SYSTEMS FOR PRODUCTION OF API 335

Cooldown of reagent
Jacketed
collection vessel

Feed
Solution #2 Plug flow reaction
in heat exchangers
Feed
Solution #1

Temperature control
Pre-heaters on jackets of heat
Mass flow
Gear pumps meters exchangers

Static mixer

FIGURE 14.13 Laboratory setup for development of the Simmons–Smith reagent continuous process.

Et2Zn + DME + CH2I2 ----► EtZnCH2I * DME + EtI Furukawa

EtZnCH2I * DME + CH2I2 ----► Zn(CH2I)2* DME + EtI Wittig

F+B W ΔH2 = − 94 kJ mol (14.7)

FIGURE 14.14 Modeled formation of complexes for the

cyclopropanating reagent.
A = Diethyl zinc/dimethoxyethane
14.8.3 Modeling and Simulation B = Diiodomethane
F = Furukawa complex
A reaction engineering approach similar to that described ear-
lier was employed here to gain further insight into the contin- W = Wittig complex
uous process. The proposed reactions for the model are Rate expressions:
formation of the Furukawa complex and Wittig complex
and are shown in Figure 14.14. A proposed kinetic model
k = Ae −E RT (14.8)
and energy balance equation governing the reaction are
shown as follows:
r1 = −k1 CA CB (14.9)
Non-isothermal Plug Flow Reaction Model
r2 = − k2 CF CB (14.10)
Assumptions

• Completely mixed in radial direction.

Simplified mass/energy balance:
• No diffusion in flow direction (axial).
• Constant shell side temperature.
• Constant stream density. dT
uρCp = ΔH1 r1 + ΔH2 r2 − UAV T − Tc (14.11)
• A two-step reaction mechanism producing first the dz
Furukawa complex followed by the Wittig complex.
• Modeled on a per tube basis in the heat exchanger.
• Overall heat transfer coefficient is independent of the u = reaction stream velocity
number of tubes in the heat exchanger (constant shell ρ = reaction stream density
side temperature) CP = reaction stream heat capacity
Reaction mechanism: T = reaction stream temperature
z = axial position in PFR
A+B F ΔH1 = − 99 kJ mol (14.6) ΔHi = heat of reaction
336 CHEMICAL ENGINEERING IN THE PHARMACEUTICAL INDUSTRY

ri = rate of reaction about 1.5 seconds, the maximum rate of heat generation
U = overall heat transfer coefficient was experienced there, at a reactor position of 0.18 m. As
AV = specific heat transfer area (area/unit volume) a consequence, the temperature of the reaction stream
increased by 18 C since the static mixer was not jacketed.
Tc = temperature of jacket coolant
The static mixer could have been jacketed for additional
In this example, the reaction kinetics were not studied in temperature control, but it was not necessary in this case.
separate detailed studies. Rather, the activation energies and The simulated reaction shows a maximum temperature of
frequency factors were fitted using the process stream tem- 75 C at about 0.27 m down the reactor. At this point, the
peratures at numerous reactor locations, a calculated overall heat generation is equal to the heat removal by the coolant
heat transfer coefficient, and information on complex for- flow. Although the predicted maximum temperature was
mation. Several assumptions were made in the modeling not measured experimentally due to limited thermocouples
of this process. Ideal plug flow was assumed although the in the PFR, the results seem reasonable and are consistent
Reynolds number was low and suggested laminar flow. with the proposed reaction mechanism and experimental
The surface temperature of the heat exchanger tubes was observations. The reaction reached 88% yield prior to being
assumed constant. Calorimetric studies provided the heat thermally quenched to less than −10 C for complex stabil-
of reaction data. The kinetic model was fitted to the temper- ity. Overall, the simulation does an adequate job in model-
ature profile with an estimated overall heat transfer coeffi- ing the observed behavior and results from the laboratory.
cient [28]. Using the kinetic parameters, the reaction was Despite limited knowledge of the reaction kinetics prior
simulated as it progressed through the reactor. Several local to modeling, the exercise demonstrates the utility of simu-
minima were determined during the model fitting exercise, lating a non-isothermal PFR. This type of process knowl-
requiring additional data to refine the model. The simulation edge could be used to modify reaction conditions if
results are shown in Figure 14.15, which plots temperature necessary but in this case was primarily used to guide the
and heat generated versus axial position in the reactor. design and operation of pilot plant and commercial manu-
Although the residence time in the static mixer was only facturing reactors.

80 2000

Model temperature
70 Jacket temperature 1800
Test 1
60 Test 2 1600
Test 3
Test 4
50 1400
Heat generation
Temperature (°C)

40 1200
Heat generation
(mW/tube)

30 1000

20 800

10 600

0 400
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8

–10 200

–20 0
Position (m)

FIGURE 14.15 Test and model results for lab-scale plug flow reactor. Note that the axial velocity is higher in the static mixer due to a lower
cross-sectional area relative to the heat exchangers. As a result, the shape of the temperature curve is influenced when plotted against position.
 PROCESS DEVELOPMENT AND CASE STUDIES OF CONTINUOUS REACTOR SYSTEMS FOR PRODUCTION OF API 337

14.8.4 Process Scale-Up to Pilot Plant 14.9 INTEGRATED CONTINUOUS

PROCESSING IN PHARMA
With little additional development work, the process
described in the previous sections was scaled up in a pilot
While implementing continuous reactions can lead to
plant to generate 700 kg of the cyclopropanating reagent
improved safety and product quality, the increased manu-
solution. The feed tanks were pressurized, and an actuated
facturing efficiencies experienced in the commodity chem-
diaphragm valve coupled with a mass flow meter controlled
ical industry are largely unrealized when the downstream
the flow rate of each feed. The process was scaled by main-
processing is conducting in a semi-batch fashion. This is
taining a similar residence time as the lab, and essentially
a result of equipment downtime in such a scenario. Cou-
numbering up the lab setup by having a larger number of
pling multiple unit operations into a continuous process
tubes in each heat exchanger, while maintaining the tube
train has the potential to accelerate introduction of new
diameter. Unlike the batch process, this approach ensured
drugs through more efficient production processes and
similar heat and mass transfer characteristics and little
decrease the costs of production with smaller facilities, min-
change in reaction conditions when moving from the lab
imization of waste, lower energy consumption, and
to the pilot plant. The PFR was constructed from a static
decreased raw material use [29]. Post-reaction processing
mixer and five shell and tube heat exchangers, each contain-
in the pharmaceutical industry typically involves extrac-
ing 163 1/8 inch tubes. The total residence time was 65 sec-
tions, solvent exchanges, crystallizations, and drying, and
onds, similar to the 51 second residence time utilized in the
technologies currently exist to perform many of these unit
lab. In the pilot plant, a spiral heat exchanger was used to
operations continuously. For example, traditional chemical
facilitate the thermal quench. The design of the pilot plant
processing equipment such as Podbielniak centrifugal
system was otherwise similar to the previously described
extractors, wiped film evaporators, and continuous crystal-
laboratory system.
lizers can perform extractions, solvent exchanges, and crys-
Table 14.5 shows the different specifications for the lab-
tallizations continuously. These devices offer not only
oratory and the pilot plant setups. The operation on pilot
higher throughput but also increase efficiencies as well,
scale was similar to the laboratory process with slight
resulting in yield improvements and less waste. Addition-
differences in measured peak process temperatures, most
ally, parallel drying trains can be setup to alternately receive
likely due to differences in heat transfer characteristics.
material from upstream continuous process trains. While
The maximum reaction stream temperature, measured after
integrated continuous processing of API is in its infancy,
the first heat exchanger in the PFR, was 56–62 C. As a
some companies have efforts underway [30], and others
result, the PFR was operated with a higher jacket temper-
are collaborating with academia to develop new technolo-
ature relative to the laboratory reactor. After exiting
gies for such purposes [2]. An integrated continuous proces-
the PFR, the reaction stream was thermally quenched
sing plant may become more common in the pharmaceutical
and collected in a jacketed 2000 L reactor for later use.
industry as technologies develop and as cost pressures rise.
The process ran until all the feed solutions were consumed.
Whether these exist as smaller plants dedicated to a single
The implemented continuous process facilitated production
product or modular multiproduct plants remains to be seen.
of 700 kg of reagent of consistent quality under reproduc-
Either way, the evolution will likely be slow given the
ible conditions.
entrenchment of existing batch processing plants and the
real, or perceived, barriers to widespread acceptance of con-
tinuous processes.

TABLE 14.5 Laboratory to Pilot Plant PFR Specifications 14.10 BARRIERS TO IMPLEMENTATION OF
Lab Pilot Plant CONTINUOUS PROCESSING IN PHARMA
DME/DEZ (g/min) 70 1358
Diiodomethane/DCM (g/min) 48 940 The barriers to continuous processing in pharma are largely
Total mass flow rate (g/min) 118 2298 historical and involve GMP documentation concerns, lack of
Volumetric flow rate (ml/min) 94 1868 experience and understanding, and an existing infrastructure
PFR residence time (s) 50 65 designed for batch processing. The pharmaceutical industry
Fluid velocity (cm/s) 1.7 3.85 has traditionally preferred batch processing largely because
Number of tubes (per shell) 19 163 of GMP documentation and traceability purposes. By having
Tube size (ID, cm) 0.254 0.254 obviously defined discrete batches or lots of material, it is
Re # 53 120 straightforward to meet GMP requirements to document
Reactor length (m) 1.0 2.55 and verify the processing activities, parameters, and raw
Heat load (W) 240 4700
materials that go into each batch. Since continuous processes