Effective Process Design & Optimization

White Paper
Real-Time FTIR and Kinetic Modeling

Table of Contents
1. Introduction 2
2. Examples of Kinetic Modeling in Real-World Investigations 4
2.1 Kinetic and Mechanistic Investigations to Enable a Key Suzuki Coupling for Sotorasib
Manufacture–What a Difference a Base Makes 4
2.2 Optimization of an Azaindazole Series of CCR1 Antagonists and Development of a
Semicontinuous-Flow Synthesis 6
2.3 Palbociclib Commercial Manufacturing Process Development. Part I: Control of
Regioselectivity in a Grignard-Mediated SNAr Coupling 8
2.4 Development of a Modeling-Based Strategy for the Safe and Effective Scale-up of
Highly Energetic Hydrogenation Reactions 10
2.5 Summary of Examples 13
3. Additional Articles Using Kinetic Modeling 14
4. Addendum 18
 1. Introduction

Investigating and understanding reaction kinetics, mechanism, and the effect of reaction variables are critical to
Unlocking Essential Process Knowledge

the development of high-yielding chemical reactions with tightly controlled impurity formation. The development
of novel syntheses in academic research and chemical and pharmaceutical process development depends on
in-depth understanding of these factors. Convenient, advanced kinetic modeling software combined with data-rich
process analytical technology (PAT) measurements gives scientists unparalleled access to fundamental kinetic and
mechanistic knowledge of complex syntheses and processes.

In the examples reviewed in this white paper, ReactIR™ PAT used in conjunction with the kinetic modeling
software packages Reaction Lab™ and Dynochem® illustrate how these powerful methodologies aid in unlocking
an in-depth understanding of synthetic chemistry. This white paper will provide a basic understanding of the value
of combining data-rich experimentation with kinetic modeling and provide examples of important
applications including:

1. Kinetic and mechanistic understanding to optimize a process using a Suzuki-Miyaura coupling reaction
2. Scale-up and process design of a Curtius Rearrangement reaction involving hazardous intermediates
3. Understanding and overcoming stalling in an SNAr reaction
4. Safe scale-up and optimization of a hydrogenation reaction

1.1. In-Situ FTIR and Kinetic Modeling

ReactIR (in-situ FTIR spectroscopy) is a versatile PAT solution used in a variety of chemistry applications including
organic synthesis, catalysis, hazardous chemistry, low or high-temperature reactions, and biosynthesis. By
providing real-time measurement under actual reaction conditions, ReactIR ensures that analytical data reflects an
in-depth and realistic view of the process, and reduces the potential negative effects of sample removal (for offline
analysis). Identifying and monitoring key reaction species provides essential experimental data for kinetic analysis.
This data helps determine reaction rates and activation energies, supports proposed mechanisms, and clarifies the
relationship between reaction variables and reaction performance.

The value of this PAT data can be enhanced using Reaction Lab—a powerful software tool used to quickly create
mechanistic models based on experimental data. Reaction Lab models the effects of variables such as temperature,
pressure, reactant loading, product and by-product concentrations. This helps to identify optimal reaction conditions
for reaction optimization and scale-up. Once a model is built, the impact of changing key reaction parameters can
be determined, providing insight into yield and impurity tradeoffs as well as the overall robustness of the reaction.
Thus, mechanistic modeling enables the exploration of design space with far less physical experimentation required.
The information from these tools facilitates a greater understanding and support for proposed reaction mechanisms,
allowing for more effective process design in a much faster time and more sustainable manner than traditional
large-sized DoE studies.

2 METTLER TOLEDO White Paper

 Benefits of Combining
ReactIR & Reaction Lab

Accessible and Easy to Learn Advances Scale-up, Tech Transfer, and Sustainable
• Transform qualitative kinetic understanding into a Manufacturing While Reducing Resource Requirements
quantitative and predictive kinetic model • Identify reaction-driving forces and critical process
• Synthetic and process development chemists can parameters using fewer, information-rich experiments
derive advanced insights without modeling specialists • Real-time data with ReactIR negates the need for the
development of analytical methods and sampling
Convenient and Data-Rich procedures that are required for offline analysis
• Obtain a rich experimental dataset in real time with • Quickly generate and validate reaction design using
ReactIR, and easily import it into Reaction Lab to the combination of experimental data from ReactIR
build a kinetic model and quantify reaction kinetics and the rapid generation of a large number of
• Detect and track reactive and transient intermediates in-silico experiments performed by Reaction Lab
with ReactIR to gain better mechanistic
understanding to develop better kinetic models Inherently Safe Operations
• Calibrated IR data provides concentration vs. time • Safely monitor and analyze hazardous reactions in
profiles that enable the determination of reaction situ with ReactIR, as opposed to offline sampling
rates, activation energies, and reaction orders • Determine safe operating conditions using kinetic
modeling with Reaction Lab

METTLER TOLEDO White Paper 3

 2. Examples of Kinetic Modeling in Real-World Investigations

Dynochem software, part of the METTLER TOLEDO Scale-up Suite, is an advanced modeling software that has been
Unlocking Essential Process Knowledge

extensively used in the chemical and pharmaceutical industries. Reaction Lab is a newer offering that provides
organic chemists with the same power of Dynochem’s kinetic modeling in a format that is exceptionally convenient
for routine use in chemical development or scale-up applications. The following articles from the peer-reviewed
literature used Dynochem, but all could have also been generated using Reaction Lab. We will demonstrate the
advantages of using Reaction Lab later in this publication, using case study 2.1 as an example.

2.1. Kinetic and Mechanistic Investigations to Enable a Key Suzuki Coupling for Sotorasib
Manufacture–What a Difference a Base Makes
Murray, J.I., Zhang, L., Simon, A., Silva Elipe, M.V., Wei, C.S., Caille, S. and Parsons, A.T. (2023). Org. Process
Res. Dev., 27, 1, 198–205. https://doi.org/10.1021/acs.oprd.2c00332

“
A key stage in Amgen’s 5-step process for manufacturing sotorasib is
the formation of a heteroaryl–aryl bond via a Suzuki–Miyaura coupling
Our mechanistic
reaction. In the work described in their article, the authors investigated
investigations into this
the kinetics and mechanism of the coupling reaction in the synthesis of
process began with the use
sotorasib to determine the catalytic pathway and rate-determining step
of reaction progress kinetic
(Figure 1).
analysis to determine the
Boc Boc effect of each component
N F HO N
F F OH on the absolute reaction
N O Pd(DPEphos)Cl2 (0.6 mol%) N
Me
N N
Cl B B
OH O B O F KOAc (2.0 eq)
Me
N
rate. Continuous reaction
+ N

O
N
F OH 2MeTHF/water, reflux
O
N F monitoring was achieved
iPr Desired iPr
N N via in situ IR spectroscopy,
Aryl chloride 1 Boroxine 2 Product 3 which enabled accurate
Monitored by ReactIR H
Protodeborylation F OH tracking of the reaction
Undesired progress with respect to
3-fluorophenol
Monitored by ReactIR the consumption of aryl
Figure 1: Sotorasib process and key species to monitor. chloride 1 and the formation
of 3-fluorophenol (the
protodeborylation byproduct
The Amgen manufacturing process for this step coupled an acid chloride
with a boroxine via a Pd(DPEphos)Cl2 catalyst and KOAc base. A major from boroxine 2)”
process challenge was controlling the formation of a protodeborylation
reaction by-product. Though the process scaled successfully, there were
several operational challenges:

• Pd catalyst had to be charged after heating to reaction temperature

• Slow addition of boroxine was required to ensure complete conversion of aryl chloride
• Incomplete conversion of aryl chloride required additional charges of catalyst and boroxine

These operational challenges reduced the efficiency of the process, increased cost and increased
by-product formation.

4 METTLER TOLEDO White Paper

Kinetic studies were performed to investigate the root cause behind the undesired pathway (protodeborylation) and
to see if it could be controlled. Reactions were performed in an EasyMax synthesis workstation equipped with in-situ
FTIR spectroscopy (ReactIR), enabling continuous tracking of the key reaction species under well-controlled reaction
conditions. Initially, kinetic studies assessed the relative sensitivity of the productive coupling and protodeborylation
to variation in concentrations of aryl chloride, boroxine, KOAc base and Pd catalyst. ReactIR measurements tracked
the consumption of the aryl chloride and formation of the 3-fluorophenol by-product. The concentration of the
boroxine reagent and the biaryl product were calculated from reaction stoichiometry and validated using HPLC.

The study revealed the driving forces behind both the product forming and impurity forming pathways. ReactIR
data showed that Impurity formation increased as boroxine increased, but was not influenced by catalyst or base
concentration (Figure 2). The product formation was only influenced by the catalyst concentration. These studies
showed that control of boroxine concentration was key to controlling the rate of impurity formation.

120 120
100 100

[Impurity] (mM)
[Product] (mM)

80 80
Figure 2: In-situ FTIR spectroscopy
60 60
shows that boroxine concentration has
40 40
Base Case (140 mM Boroxine) Base Case (140 mM Boroxine) no impact on product formation (left
20 Reduced Boroxine (100 mM) 20 Reduced Boroxine (100 mM) plot), but has a first-order effect on
Increased Boroxine (180 mM) Increased Boroxine (180 mM)
0 0 formation of the 3-fluoroaryl impurity
0 5 10 15 20 0 5 10 15 20
(right plot).
Time (hrs) Time (hrs)

These results rationalized reducing the boroxine dosing rate to minimize by-product formation. The addition of extra
catalyst, along with the reduced boroxine dosing rate, mitigated catalyst deactivation associated with the formation
of the by-product.

The kinetic studies led the authors to discover another significant enhancement to the process: substituting K2CO3 for
KOAc. Initial kinetic studies showed that boroxine and the catalyst were the only driving forces that could be used to
control the reaction. However, the model revealed that controlling these two factors alone would not give the desired
outcomes. This led the authors to explore other options with more factors to control. They found that changing the
base to K2CO3 gave a different mechanism. The carbonate base caused the reaction to proceed through first-order
kinetics with respect to the boroxine. Changing the base also changed the rate-determining step from reductive
elimination to transmetalation, which significantly improved selectivity for the desired pathway and reduction of the
fluorophenol by-product.

In the original article, the reactions were modeled using Dynochem

software using an “All-In” batch model. The fitting process
used the data for the standard reaction, the boroxine different
excess data (125%), and reduced temperature data to refine the
“ The plethora of kinetic data
collected during the course of
these studies was utilized in the
predicted reaction rates and to test the effect of these variables construction of a predictive model
on the reaction. The data showed an excellent fit under a range of for the K2CO3 process…Excellent
conditions to the following kinetic model: agreement with the experimental
data collected during our kinetic
Aryl Cl 1 + Pd cat ► Pd-Aryl studies was observed. This
Pd-Aryl + Boroxine ► Boroxine-Pd-Aryl (RDS) model was then deployed to
Boroxine-Pd-Aryl ► BiAryl 3 + Pd determine the minimum catalyst
Boroxine 2 + H2O ► 3F-Phenol + B(OH)3 and boroxine loadings required to
achieve the desired conversion.”

METTLER TOLEDO White Paper 5

 Yield Product
Kinetic model validation required just four experiments
0.22 100 based on various conditions across the described chemical
0.2 space (Figure 3).
0.18
Unlocking Essential Process Knowledge

90 Impact of combined use of in-situ IR and

Catalyst Loading

0.16
kinetic modeling
0.14

0.12
80
The enhanced mechanistic understanding from the in-situ
0.1
IR measurement and kinetic modeling resulted in several
process improvements, including the use of a bulk charge
0.08

70
of boroxine rather than slow addition, a reduction in the
0.06
boroxine charge from 0.50 eq to 0.40 eq, and of the Pd
1 1.05 1.1 1.15 1.2
Boroxine Equivalents
loading from 0.6 to 0.2 mol%. The optimized process was
more robust and cost-efficient (consuming less catalyst
Figure 3: The kinetic modeling results were validated using just
four experiments based on conditions across the chemical space. and boroxine) and simpler to perform at scale.

2.2. Optimization of an Azaindazole Series of CCR1 Antagonists and Development of a

Semicontinuous-Flow Synthesis.
Harcken, C., Grant, J., Razavi, H., Marsini, M.A., Buono, F.G., Lorenz, J.C. and Jonathan T. Reeves. (2019).
Complete Accounts of Integrated Drug Discovery and Development: Recent Examples from the Pharmaceutical
Industry Volume 2, 8, 185-238. ACS Symposium Series Vol. 1332.
https://doi.org/10.1021/bk-2019-1332.ch008

“
There is significant interest in the development of chemotactic cytokine
receptor 1 (CCR1) antagonists in order to treat inflammatory and
…mechanistic guidance via
autoimmune disease. In a program to develop CCR1 antagonists,
PAT and in-house process
a novel series of azaindazole compounds was synthesized. One
development of continuous-
compound, a cyclopropyl benzylamine-substituted 6-azaindazole,
flow technology allowed for
showed promise, and the preclinical development of this compound
the execution of a concise
was undertaken to facilitate a large-scale synthesis. Different synthetic
and scalable synthesis of
routes were assessed, but ultimately a Curtius rearrangement
CCR1 antagonist 60… critical
approach was chosen for development – despite this being a
mechanistic understanding
less-desirable transformation to run on a large scale. The Curtius
of each process parameter
rearrangement would provide access to a carbamate that could
allowed for accomplishment of
be deprotected to form an amine salt, which could then be used in
the first example of direct amide
an amide coupling. There was significant concern in the choice of
synthesis semicontinuous
this strategy due to the formation of toxic and highly reactive acyl
Curtius rearrangement and
azide and isocyanate species. To ensure safety and quality, in-depth
acid–isocyanate coupling as
mechanistic understanding and safety testing was required before
the key step in a highly efficient
scale-up could be undertaken.
and an environmentally friendly
API synthesis.”
To pursue the goal of mechanistic understanding, in-situ FTIR
(ReactIR) was used as the process monitoring technology. The
starting carboxylic acid 1 was reacted with diphenylphosphoryl azide
(DPPA) as the azide source, first forming the acyl azide 2. This intermediate underwent thermal rearrangement to
isocyanate 3, and was then captured as the carbamate 4 by reaction with t-amyl alcohol (Figure 1).

6 METTLER TOLEDO White Paper

 O O
N
H
N O
Time-course IR allowed for tracking starting
OH DPPA, Et3N N3 ?,-N2 C t-AmOH
O O material, DPPA (2168 cm-1), as well as the unstable
SO2Me SO2Me
N SO2Me N N N SO2Me
intermediates, azide 2 (2147 cm-1) and isocyanate 3
1 2 3 4
(2271 cm-1) (Figure 2).
Figure 1: Reaction scheme showing azide and isocyanate intermediates.

100 The main focus of the IR experiment was to gain an

2 understanding of the kinetics of the formation and
80 degradation of the intermediate azide and isocyanate
Normalized Peak Height %

DPPA
3 compounds. The IR data showed that formation of
60
the first intermediate (the azide 2), was rapid. The
40
subsequent reaction the isocyanate (3) was slow,
and lead to a build-up of the toxic and potentially
20 explosive intermediate 2. Further studies with the
ReactIR showed that changing temperature alone
0
0 2 4 6 could not prevent the buildup of this intermediate, so
Time (hours) an alternate control strategy was developed. Slow
Figure 2: ReactIR allows for tracking of both starting material 1, as well addition of DPPA at elevated temperature was used
as unstable intermediates 2 and 3.
to keep the concentration of the azide low.
O
1, -CO2 N N An initial kilogram-scale run of the reaction resulted
N H H
C
O
urea formation N N
in lower yields than anticipated, resulting from
SO2Me 5 SO2Me
the formation of two major impurities that were
N SO2Me O
characterized as urea 5 and amide 6. It was
N
3 3, -CO2
H
N
suspected that impurity 6 was formed by reaction
amide formation
SO2Me
of the isocyanate with the acid starting material 1,
MeO2S N
6
while impurity 5 was formed by dimerization of the
Figure 3: Proposed mechanism of formation of urea 5 and amide 6 isocyanate 3 (Figure 3).
impurities.

Using in-situ FTIR measurements, kinetic analysis

showed that the formation of these impurities resulted from the excess acid 1 in the presence of isocyanate 3,
which occurred during slow DPPA addition on scale-up. This reinforced the importance of the DPPA addition rate in
minimizing impurity formation.

In order to determine the best

Impurity Starting Material (1) Product (4)
addition rate to minimize impurity,
(a) 2.0 (b) 2.0
a kinetic model was developed
1.6 Slow addition DPPA 1.6 Rapid addition DPPA
based on the FTIR data. This
Process Parameters

60% product yield 80% product yield

1.2 1.2
model accurately predicted a
yield of ~60% for addition of 0.8 0.8

DPPA over 20 minutes on lab 0.4 0.4

scale (Figure 4a), and a yield 0.0

0.0 17.5 35.0 52.5 70.0 87.5
0.0
0.0 17.5 35.0 52.5 70.0 87.5

increase to ~80% by reducing Time (min) Time (min)

Figure 4: Effect of DPPA dosing rate on carbamate product yield.

the addition time to 2 minutes
(Figure 4b).

METTLER TOLEDO White Paper 7

 Although rapid addition of DPPA was feasible on a lab scale, it was very challenging to perform under semi-
batch conditions at a larger scale. A CSTR/batch approach was considered; however, this was discontinued due
to concerns about the build-up of hazardous intermediates. Instead, based on the mechanistic understanding
and simulations obtained using ReactIR data with kinetic modeling, a high-temperature continuous-flow solution
Unlocking Essential Process Knowledge

was implemented. This enabled the DPPA and acid 1 to be mixed simultaneously so that the rearrangement and
carbamate formation occurred in the absence of unreacted 1, and impurity formation was inhibited (Figure 5).
Additionally, the smaller reactor size and better control of the reagent stream resulted in a safer process.

OH

O
H
N N OR
N SO2Me N3 C
∆ O ROH
O
acid 1 fast slow slow

N SO2Me N SO2Me N SO2Me

DPPA
azide 3 isocyanate 3

Feed A:
acid 1, NEt,
PMBOH,
Jacketed Plug-flow N2
Reactor (135ºC)
toluene
Pump A Inline
FTIR
Back-
pressure
regulator
T-mixer Recieving Tank:
Feed B: Carbamate
DPPA/ Pre-heated
toluene Jacketed
Pump B
Static Mixer
Modular Flow Reactor System

Figure 5: Continuous flow strategy based on kinetic analysis.

Using optimized conditions, a throughput of 0.75 kg acid 1 per hour provided a 90% assay yield and ~80%
isolated yield of the carbamate 4. Inline FTIR was used to ensure no accumulation of the acyl azide or
isocyanate intermediates.

Impact of combined use of in-situ IR and kinetic modeling

Real-time IR data used in modeling simulations resulted in an in-depth understanding of the reaction mechanism
and kinetics. It was determined that continuous flow technology operating at elevated temperature with
simultaneous reagent mixing led to higher yield, inhibited impurity formation and provided an overall safer process.

The implementation of this continuous flow strategy was successfully executed on a 40-kg scale to support
clinical trials.

2.3. Palbociclib Commercial Manufacturing Process Development. Part I: Control of

Regioselectivity in a Grignard-Mediated SNAr Coupling.
Duan, S., Place, D., Perfect, H.H., Ide, N.D., Maloney, M., Sutherland, K., Price Wiglesworth, K.E., Wang, K.,
Olivier, M., Kong, F., Leeman, K., Blunt, J., Draper, J., McAuliffe, M., O’Sullivan, M. and Lynch, D. (2016). Org.
Process Res. Dev., 20, 1191−1202. http://dx.doi.org/10.1021/acs.oprd.6b00070

As an inhibitor of cyclin-dependent kinases 4 and 6 (CDK4/6), the drug palbociclib is used to block tumor cell
growth in specific cancers. The initial process route was scalable but not optimal for a variety of reasons. Among
these, a key step involving an SNAr coupling required nearly two equivalents of starting material. This starting

8 METTLER TOLEDO White Paper

“ Two online analytical tools were utilized to collect the required information. First, in situ ReactIR
monitoring of the reaction progress provided some insight into the reaction mechanism. Based on the IR
data, deprotonation of aminopyridine was instantaneous upon addition of LiHMDS… Subsequent addition
of pyrimidine resulted in the formation of a small amount of desired product and a larger amount of an
intermediate…Using a Mettler Toledo EasySampler probe, a tool that can take multiple samples in short
period, we were able to capture samples containing this intermediate.”

material required several synthetic steps Me

Br
to produce, and higher equivalents weak acid
NH2 N

significantly increased manufacturing weak base N N N

R
O
flourine mediated
costs. As a result, development work NH2 pd mediated
N
focused on optimizing the coupling N
Me
undesired pathway
Br
N
reaction and identifying conditions to N
R Me
Cl N N O
ensure process robustness. N
R N
Br

N HN N N O
R R
strong base
The SNAr reaction between the amino pyridine 1 pyrimidine compound 2
N

amino pyridine 1 and the pyrimidine desired product

N
Issue: reaction stalls at 50%
compound 7 (Figure 1) involves a
N
relatively complicated mechanism, and R

most approaches involved reaction with Product 4

the endocyclic pyridine nitrogen atom Figure 1: Strong base provides product, but why does the reaction stall?
(green) rather than the desired exocyclic
Me
amine (blue). By using LiHMDS as a Me Me
N
Br
Br Br
strong base, the correct regioselectivity NH2 NH N NH2 N
HN N N
R
O
N N Cl N N O N + N N O
was achieved; however, the reaction R N
R
LiHMDS

stalled at 50% yield. N N N

Compound 2 N

N N N
R N
R R R
Compound 1 Deprotonated 1 Intermediate 3 Product 4
The initial hypothesis was that the Detected using ReactIR
capture using EasySampler
LiHMDS base was being consumed by
2.0
the product, preventing further reaction. Product 4

To test this assumption, the reaction 1.5 Deprotonated 1

was investigated using in-situ ReactIR Compound 1

measurements.
ConclRT

1.0

Intermediate 3

ReactIR provided insight into the 0.5

reaction mechanism, indicating that

0.0
deprotonation of aminopyridine 1 was 00:00 00:15 00:30 00:45 01:00 01:15 01:30 01:45: 02:00
Time
instantaneous upon addition of LiHMDS Figure 2: ReactIR tracks key species including intermediate.
(Figure 2). A small quantity of product
4 was formed with the subsequent addition of compound 2, but a larger quantity of intermediate 3 also formed and
slowly converted to product 4.

Using an EasySampler™ probe, reaction samples that contained the intermediate were collected, enabling its
identification as the pyridinum salt 3 (Figure 2).

METTLER TOLEDO White Paper 9

 Kinetic modeling was performed to fully understand Experimental Data Model Data
Compound 1 Compound 1
the reaction mechanism of the intermediate. The 0.25 Compound 2 Compound 2
Intermediate 3 Intermediate 3
modeling supported a first-order reaction for the Product 4 Product 4
0.20
conversion of an imino intermediate 3 to product 4

Concentration (mol/L)
Unlocking Essential Process Knowledge

(Figure 4) – this was confirmed by additional ReactIR 0.15

experiments, which also showed that the decay of the
0.10
intermediate was first order.
0.05

This suggested a new mechanism in which

0
the intermediate reacted via an intramolecular 0 20 40 60 80 100 120

rearrangement mechanism (Figure 4), not a

Figure 3: Modeling indicates first-order kinetics for imino intermediate
bimolecular reaction as expected. The model results conversion to product.

Starting Material 1 Starting Material 2

Me Br O
NH2 Br RN N NH
Me N N N NR
N N
Br LiHMDS
N HN N N O N
R Excess RN Me N N
N Cl N N O N
R HN N NR
O Br
N Product 4
N
R Dimer Impurity 5
Leads to reaction stalling
Me N Controlled through base selection
R
Br
NH2 N
Rearrangement
LiHMDS N N N O Confirmed with
R kinetic model

N
Intermediate 3
Seen by ReactIR
N
R
Figure 4: Based on IR data and modeling, an intramolecular rearrangement mechanism is proposed to be responsible for the reaction stalling.

suggested that another pathway was responsible for the stalled reaction. Further investigation revealed that a
dimeric impurity 5 was formed, consuming an additional equivalent of base, which stalled the reaction. It was
found that this impurity could be controlled through adjusting the pKa of the base. Ultimately, changing the base to
CyMgCl overcame the reaction stalling and lead to a more sensible, scalable process.

Impact of combined use of in-situ analysis and kinetic modeling

In-situ FTIR measurements indicated that an intermediate formed upon addition of LiHMDS—in-situ sampling and
offline HPLC identified the intermediate. Kinetic modeling provided insight into the mechanism of the intermediate
formation, and the presence of a dimeric impurity. Further investigation determined that undesired dimer formation
could be controlled in situ through modification of the base pKa. A modified manufacturing process was developed
using the Grignard base CyMgCl in lieu of LiHMDS. This new process using the Grignard base mitigated the reaction
stalling by fully preventing the dimer formation pathway and enabling the desired increase of product yield.

2.4. Development of a Modeling-Based Strategy for the Safe and Effective Scale-up of
Highly Energetic Hydrogenation Reactions
Mitchell, C.W., Strawser, J.D., Gottlieb, A., Millonig, M.H., Hicks, F.A. and Papageorgiou, C.D. (2014). Org.
Process Res. Dev., 18, 1828−1835. https://doi.org/10.1021/op500207r

The catalytic reduction of aromatic nitro groups to their corresponding anilines in the presence of H2 is widely used
in pharmaceutical manufacturing. In order to scale-up from development to manufacturing scale and to ensure
the efficiency, quality, and safety of this highly exothermic reaction, it is necessary to understand a number of key

10 METTLER TOLEDO White Paper

 “
variables. For example, mixing performance controls how
quickly H2 is dissolved in solution. The authors present the …the highly exothermic hydrogenation
example of a highly exothermic, heterogeneous hydrogenation of a substituted nitropyrimidine can be
of nitropyrimidine 1 (Figure 1) that was previously scaled up to successfully modeled by performing
11-kg scale in a 200 L hydrogenator. a small number of reactions and
taking advantage of in-line process
NO2 NH2
N
H2, 30 psi
N analytical tools (FT-IR and H2 gas
CO2Et CO2Et
Cl N N
F F Pt/C (catalyst)
Cl N N
F F uptake) to collect detailed reaction
profile data without necessitating the
1 2 tedious analysis of R samples. Utilizing
Figure 1: Catalyzed hydrogenation of nitropyrimidine. Scale-up Systems’s DynoChem 2011,
it has been possible to use this data
to create both a comprehensive kinetic
Despite maximum cooling in manufacturing, the hydrogen model and a simplified physical
dosing system was unable to keep up with demand, model capable of predicting reaction
requiring periodic adjustments of several process variables. temperature profiles as well as the
To understand and improve reaction scale-up, a predictive maximum expected exotherm, as a
computational model that included key variables such as function of agitation speed.”
temperature, pressure, substrate loading, catalyst loading, and
agitation rate was developed.

An RC1 calorimeter equipped with a 1 L pressure vessel was used for hydrogenation of nitropyrimidine 1. In addition
to a temperature probe, the vessel was equipped with an inline FTIR probe (ReactIR) for time-course tracking of key
reaction species.

100 A key part of developing hydrogenation processes is

determining the impact of stir speed. Under poor mixing
80 800 rpm [0.4 s-1] conditions, the reaction is “hydrogen starved”, as hydrogen
1600 rpm [1.2 s-1]
Nitropyrimidine, 1 (%)

2000 rpm [1.5 s-1] uptake is much slower than the catalytic reaction. Only with
60
sufficient mixing will hydrogen be dissolved quickly enough
40
that the intrinsic kinetics of the system can be studied. The
in-situ IR measurements enabled quick determination of the
20 mixing conditions needed for studying and understanding the
kinetics of the system (Figure 2).
0
0 5 10 15 20 25 30 35 40

The in-situ IR data also helped identify the correct reaction

Figure 2: Tracking nitropyrimidine reagent consumption rate v.
Time (minutes)

agitation rate. mechanism. The IR data did not reveal the presence of
potential intermediates such as hydroxylamine, which
implied that the rate-limiting step was the initial reduction of the nitropyrimidine to a nitroso intermediate. However,
investigations with different H2 pressures suggested that the initial reaction rate did not follow first-order kinetics with
respect to H2 concentration, indicating a more complex reaction mechanism was occurring involving adsorption
onto the catalyst surface. This was further confirmed when attempting to fit a first-order kinetic model based on the
available analytical data. Although good fits were found for any one data set, the model could not predict the test
reaction outcome when conditions were varied.

Instead, a kinetic model based on ReactIR data was built that showed that a Langmuir-Hinshelwood mechanism
was taking place, indicating that the relative binding of each species to the catalyst is what drives the reaction
(Figure 3).

METTLER TOLEDO White Paper 11

 This model accounts for adsorption of species on catalyst 0.50 1st Order Kinetic Model
0.45
surface sites at a rate proportional to both the partial 0.40
pressure or concentration of the species, and the number of 0.35
0.30
free active sites available. The adsorbed species then form

Moles
0.25
Unlocking Essential Process Knowledge

adsorbed products at a rate related to their concentrations 0.20

0.15
on the surface. 0.10
0.05
0
Impact of combined use of in-situ IR and kinetic modeling 0 100 200 300 400 500
Time (minutes)
Kinetic data acquired from ReactIR measurements
0.50 Langmuir Hinselwood Model
performed in an RC1 pressure vessel led to understanding 0.45

the effect of reaction variables such as mixing, as well as 0.40

0.35
insight into the mechanism. A kinetic model based on the 0.30

Moles
in-situ IR elucidated the mechanism for the driving force of 0.25
0.20
the reaction. With an understanding of this mechanism in 0.15
0.10
hand, the kinetic model was extended by adding enthalpy 0.05
values from literature for each reaction step to enable the 0
0 100 200 300 400 500
prediction of heat output and the anticipated exotherm Time (minutes)

for a scaled-up reaction. The model was then expanded Experimental Model Prediction
Nitropyrimidine Product H2 uptake Nitropyrimidine Product H2 uptake
to aid in equipment selection, allowing for large reactors
Figure 3: Kinetic modeling supports a Langmuir-Hinshelwood
to be chosen which have adequate mixing, dosing, and mechanism, where the relative binding of each species to the
heat exchange for safe and high yielding scale-up. This catalyst is what drives the reactionadsorption mechanism.
enhanced model was then validated by successfully
predicting the performance of a 7x scale-up to produce
350g of product.

12 METTLER TOLEDO White Paper

2.5. Summary of Examples

The case studies demonstrate that using PAT methods such as ReactIR in conjunction with kinetic modeling
software such as Reaction Lab provides unparalleled insight into reaction kinetics, mechanisms, and the effect
of variables on reaction outcome. Kinetic modeling with experimental data minimizes the number of physical
experiments required to explore “what if” scenarios and greatly accelerates process development and optimization.

Catalyzed Coupling Curtius Rearrangement SNAr Hydrogenation

Primary Issue Impurity Control Scalability Stalled Reaction Safety
In-situ monitoring allowed Kinetics of Intermediate formation Intermediate formation Kinetics in a
measurement of… impurity formation high-pressure system
Kinetic modeling A change in catalyst Intermediate degradation Dimer formation Scale-up reaction rate
demonstrated that… resting state was needed could be prevented by consumed reagent and and heat output could
rapid addition stalled reaction be predicted
Leading to… Better impurity Transition to a successful New conditions to Safe and successful
control and more 40 kg flow process overcome reaction scale-up of the reaction
efficient process stalling and deliver a
more sustainable process

ReactIR monitoring of key reaction species provides… ReactIR data with Reaction Lab modeling enables…
• by-product formation • Informed decision making on the optimum reaction
• starting material degradation technology selection for your system (e.g. batch,
• catalyst deactivation fed-batch, PFR, CSTR, etc.)
• catalyst inhibition • Understanding of parameters critical to control of
your chemistry
• Prediction and control of impurity formation
• Robust prediction of scale-up outcomes using only
lab-scale data

METTLER TOLEDO White Paper 13

 3. Additional Articles Using Kinetic Modeling

Wang, J., Agrawal, P., Buser, J.Y., Embry, M.C., McFarland, A.D., Berglund, M.R., Groh, J.M. and Viswanath,
Unlocking Essential Process Knowledge

S.K. (2023). Development of a Mechanistic Amino Acid Activation Reaction Kinetics Model for Safe and
Efficient Solid Phase Peptide Synthesis. Indus. Eng. Chem. Res., 62, 37, 14986–14996.
https://doi.org/10.1021/acs.iecr.3c02145

“Amino acid pre-activation is a critical reaction step during solid phase peptide synthesis due in part to the
formation of impurities such as isomer, N,N′-diisopropylcarbodiimide-adduct, single amino acid addition, and
the generation of hazardous hydrogen cyanide side product… we present a detailed mechanistic amino acid pre-
activation kinetics model to correlate the input parameters with the output reaction conversion, isomer formation,
and hydrogen cyanide generation response…Reaction kinetics parameter fittings, full factorial design of experiment
and sensitivity analysis were performed using Reaction Lab software from Scale-up Systems. The literature data
used for kinetics parameters fitting includes Leucine pre-activation and Oxyma/DIC activation data.”

Marzijarani, N.S., Fine, A.J., Dalby, S.M., Gangam, R., Poudyal, S., Behre, T., Ekkati, A.R., Armstrong, B.M.,
Shultz, C.S., Dance, Z.E.X. and Stone, K. (2022). Manufacturing Process Development for Belzutifan, Part 4:
Nitrogen Flow Criticality for Transfer Hydrogenation Control. Org. Process Res. Dev. 26, 3, 533–542.
https://doi.org/10.1021/acs.oprd.1c00231

“…belzutifan features a highly functionalized indane…In the clinical supply process, intermediate 3 was prepared by
asymmetric transfer hydrogenation… it was deemed unsuitable as-is for commercial manufacturing… In keeping
with our commitment to green and sustainable manufacturing, we sought to identify sustainable solutions…we
report the commercial manufacturing process for intermediate 3… The process modeling platform DynoChem was
used to simulate the rates of the process: reaction, mass transfer of CO2, and introduction/removal of headspace
nitrogen… the scalability assessment achieved through dynamic VLE modeling in DynoChem shows that the CO2
removal rate is governed almost entirely by the incoming nitrogen flow rate…PAT was employed at large scales to
investigate the reduction/deprotection step to ultimately provide online in situ reaction monitoring, detailed process
understanding, and a process fingerprint to assess the success of technology transfer to our manufacturing site. It
was shown that both infrared (FTIR) and Raman spectroscopies, combined with automated sampling and off-line
analysis, provide multiple clear finger-prints to track the reaction progress.”

Laraman, F.J., Fisk, H., Whittaker, D.T.E., Cherryman, J.H. and Diorazio, L.J. (2022). Investigating the
Activation Kinetics of Phosphoramidites for Oligonucleotide Synthesis. Org. Process Res. Dev., 26, 3, 764–772.
https://doi.org/10.1021/acs.oprd.1c00195

“AstraZeneca has an interest in the coupling of constrained ethyl (cEt) phosphoramidites… we were keen to
optimize their use in ASO [antisense oligonucleotides] synthesis…Using both in-line mid-IR spectroscopy and rapid
NMR spectroscopy, we studied the rates of activation of traditional deoxyribonucleoside phosphoramidites and
their chemically modified constrained ethyl (cEt) ribose locked nucleic acid analogues…The IR was detecting the
formation of [intermediate] 5 and NMR the formation of [intermediate]7. By proving that these two formation rates
were equivalent, we could safely assume that we were measuring the same conversion…Each reaction profile was
fitted to eq 2 using Dynochem to give a pseudo-first-order rate constant…”

14 METTLER TOLEDO White Paper

Yang, C., Feng, H. and Stone, K. (2021). Characterization of Propionyl Phosphate Hydrolysis Kinetics by
Data-Rich Experiments and In-Line Process Analytical Technology. Org. Process Res. Dev. 25, 3, 507–515.
https://doi.org/10.1021/acs.oprd.0c00451

“…enzymatic phosphorylation is used as an important step for the synthesis of an active pharmaceutical ingredient.
Propionyl phosphate (PrP) was chosen as the phosphate donor because of the benefits for the downstream
biocatalytic process…Understanding of the PrP degradation kinetics is of great importance for risk mitigation… the
hydrolysis reaction of propionyl phosphate monoammonium salt was carried out with a modified repeated

temperature scanning experiment and closely monitored by real-time IR spectroscopy calibrated to off-line NMR
analysis to obtain rich concentration data assisted by multivariate IR modeling. The concentration and reaction
temperature profiles were fit to a first-order kinetic model with valid statistical indications by the DynoChem model,
reporting two key kinetics parameters of PrP hydrolysis for the first time.”

Schnell, S.D., González, J.A., Sklyaruk, J., Linden, A. and Gademann, K. (2021). Boron Trifluoride-Mediated
Cycloaddition of 3-Bromotetrazine and Silyl Enol Ethers: Synthesis of 3-Bromo-pyridazines. J. Org. Chem., 86,
17, 12008–12023. https://doi.org/10.1021/acs.joc.1c01384

“…we report in full detail the synthesis of polyfunctionalized pyridazines by using the Lewis acid mediated inverse-
electron demand Diels-Alder (iEDDA) reaction of 3-Br-Tet (1) with silyl-enol ethers under very mild reaction
conditions. Additionally, the scale-up of the presented method and downstream diversification via various cross-
coupling reactions are demonstrated…”

Ramisetty, K.A., Kumar, K.V. and Å Rasmuson, Å, C. (2019). Advanced Size Distribution Control in Batch
Cooling Crystallization Using Ultrasound. Org. Process Res. Dev. 2019, 23, 5, 935–944
https://doi.org/10.1021/acs.oprd.9b00033

“…we propose a method that will improve control over the product crystal size distribution and will facilitate reliable
upscaling of sonocrystallization. This method relies on a two-stage crystallization process operating in batch mode,
where a sonicated small-scale crystalliser is coupled to a larger conventional cooling crystalliser. A suspension
with high quality fine sized seeds is generated using ultrasound in the small crystallizer. This suspension is then
transferred to the large crystallizer, to allow the seeds to grow under controlled conditions and thus obtain product
crystals with a controlled size and a narrow size distribution…process data obtained by PAT tools were used
together with a commercial simulation software to determine crystallization kinetics…To assist in the evaluation of
the results population balance modeling was performed in MS Excel using the DynoChem software.”

McMillan, A.E., Steven, A., Ashworth, I.W., Mullen, A.K., Chan, L.C., Galan Espinosa, M.R., Pilling, M.J., Raw,
S.A., and Jones, M.F. (2019). Stereoretentive Etherification of an α-Aryl-β-amino Alcohol Using a Selective
Aziridinium Ring Opening for the Synthesis of AZD7594. J. Org. Chem., 84, 8, 4629–4638.
https://doi.org/10.1021/acs.joc.8b01062

“In order to develop a highly convergent synthesis of AZD7594, an efficient and selective method of aryl
etherification using the readily accessible α-aryl-β-amino alcohol 3 was required (Scheme 1)… Previously
established stereoretentive methods of etherifying α-aryl-β-amino alcohols were found to be unsuitable for the
manufacture of AZD7594; instead, a double-inversion approach using an aziridinium species was successfully
developed… The mesylation of 26 was monitored with in situ attenuated total reflectance Fourier transformed

METTLER TOLEDO White Paper 15

 infrared (ATR-FTIR)… ATR-FTIR spectroscopy was also used to observe the consumption and quenching of
methanesulfonyl chloride…To address concerns over the mixing of a biphasic system on plant scale, the modeling
tool DynoChem was used to scale the power per unit mass used by the impeller of a 2 L vessel up to a
1000 L vessel.”
Unlocking Essential Process Knowledge

Hunter, S.M., Susanne, F., Whitten, R., Hartwig, T. and Schilling, M. (2018). Process design methodology for
organometallic chemistry in continuous flow systems. Tetrahedron, 74, 25, 3176-3182.
https://doi.org/10.1016/j.tet.2018.04.020

“A systematic approach using intrinsic process understanding and in-silico enabled design has been shown to
be an effective tool in the development of organometallic flow processes. Following a workflow encompassing
kinetic and calorimetric understanding the authors have demonstrated how modelling of organometallic reactions
can be used to facilitate industrial process design, which was then be demonstrated at the laboratory scale…
The differential equations for each transformation and side re-action were integrated into a series of ideal Plug
Flow Reactor model (PFR) in DynoChem, linking the two transformations into one in-silico model. For each
transformation, the rate constant, the activation energy and the enthalpy of reaction are included. In addition to the
mechanistic description, physical rates specific to the capillary PFR have been included into the DynoChem model.
This includes micro mixing time and heat transfer coefficients. These physical rates were either estimated using
mathematical expressions when sensitivity was low or were measured through a set of experiments as described
above if it was revealed as being a key parameter.”

Susanne, F., Martin, B., Aubry, M., Sedelmeier, J., Lima, F., Sevinc, S., Piccioni, L., Haber, J., Schenkel, B. and
Venturoni, F. (2017). Match-Making Reactors to Chemistry: A Continuous Manufacturing-Enabled Sequence to a
Key Benzoxazole Pharmaceutical Intermediate Org. Process Res. Dev., 21, 1779−1793.
http://dx.doi.org/10.1021/acs.oprd.7b00254

“We therefore initiated the process research and development toward compound [benzoxazole] 1 and herein outline
the transition from batch to flow, where appropriate, to achieve a more robust process with increased yield… In
order to attribute the appropriate reactor to each chemical step we began by evaluating the heat-release profile of the
sequence with reaction calorimetry in batch. The entire chemical sequence was run in an RC1 calorimeter in order to
accurately measure the heat flow during each rapid dosing step… the concentration profile of each intermediate was
followed by online FT-IR analysis of characteristic signals… Having determined the kinetics for each of the byproduct
pathways, summarized in Figure 6, these pathways were incorporated into the Dynochem-based kinetic model to
enable an overall simulation of the reaction within the flow reactor equipment.”

Changi, S.M. and Wong, S-W. (2016). Kinetics Model for Designing Grignard Reactions in Batch or Flow
Operations. Org. Process Res. Dev., 20, 2, 525–539. https://doi.org/10.1021/acs.oprd.5b00281

“The paper describes a “fit for use” model to obtain kinetics free of mass transfer and help in scale up for Grignard
reactions using either a batch or continuous mode of operation. The mass transfer rates depend on the agitation
speed and magnesium type used… The batch reactions were performed in a Mettler-Toledo RC1e setup…The
continuous experiments were run in two different size of reactors…For analyzing the reaction progress and getting
useful kinetics information, IR and HPLC tools were used. Reaction progress was quantified in situ using FTIR…
DynoChem (version 2.0) was used to determine the model parameters. Dynochem fits reaction rate constant using
Arrhenius functionality (eq 7) at a reference temperature (chosen as 0 °C for this study).”

16 METTLER TOLEDO White Paper

Drexler, M.T., Foley, D.A., Ward II, H.W. and Clarke, H.J. (2015). IR and NMR Reaction Monitoring Techniques
for Nucleophilic Addition Reactions: In Situ Monitoring of the Addition of Benzimidazole to a Pyridinium Salt.
Org. Process Res. Dev., 19, 9, 1119–1127. https://doi.org/10.1021/acs.oprd.5b00029

“In situ IR and online NMR were used to investigate the two stage reaction of benzimidazoles 3 lithiation and
subsequent addition to the pyridinium triflate salt 2, a transformation which was low yielding in initial development
experiments. …The IR data from the four experiments was converted to concentration profile… These profiles were
used to perform a fit of a basic kinetic model… The model was approached in a few different ways – the key was
to look at how the data obtained at various pyridinium salt 2 stoichiometries fit relative rate scenarios. The initial fit
of the data to the model was to assume that the rate of the first reaction, r1, was greater than the rates of the other
two reactions. The initial fit also assumed that the rate of the desired addition, r2, was faster than the rate of the
undesired quench of the anion, r3 (i.e., k2 > k3); this would present a case that the desired reaction should be
favored. Using these assumptions starting point, the data was fit using DynoChem in simulator mode.”

METTLER TOLEDO White Paper 17

 4. Addendum

4.1. Key Information Provided by Reaction Lab Modeling Using Data-Rich ReactIR
Unlocking Essential Process Knowledge

Measurements: Suzuki Coupling Reaction

The intuitive Reaction Lab model building interface is designed to be accessible to chemists, allowing them to build
their reaction model simply and rapidly. First the user defines the overall reaction and the materials involved in the
reaction. The different sections of the model are linked, so an edit to a material in one section will be automatically
updated elsewhere in the model.

Materials involved in the model can be

entered from chemical structure drawing
software, internal and external databases,
or simply based on chemical formulae.

The next stage of model building is the entry of the proposed reaction mechanism, using a simple drop-down menu
based interface. Adding heterogeneous phases or feed tanks is also quick and simple, and user-defined calculations
can be included very easily.

The Reaction mechanism is entered using

a drop-down menu.

The Process Scheme allows simple definition

of multi-phase and fed-batch systems.

18 METTLER TOLEDO White Paper

Finally, reaction starting conditions can be easily tabulated, and experimental data imported rapidly from Excel
tables. iCIR 7.2 can automatically generate suitable tables of IR data. Simple and interactive plots are generated for
each experimental data set, and plotting options enable comparison of initial reaction rates and several RPKA-type
data views.

The Data Chart plots the experimental

data. Users can select specific experiments
or analytes. Plots can have linear, log,
differential, reciprocal, or normalized scales.

The “How to” for Successful Kinetic Studies

• Design experiments to probe influence of • Use model to determine driving forces and design an
reaction variables optimized process
• Study reaction using ReactIR data • Use model to explore design space around
• Process and calibrate data, import into a optimized values
kinetic model • Validate predictions using further experimentation
• Develop and test mechanistic hypothesis by fitting with ReactIR
experimental data to the model
• Revise mechanistic hypothesis directly in the software
until a model fit is found

Fitting the data then shows whether experimental data is in agreement with the proposed mechanism. The output is
reaction rate constants, k, activation energies, Ea and fitting statistics.

Fitting gives a graphical plot of the predicted reaction profiles,

experimental data, predicted kinetic parameters and fitting statistics.

METTLER TOLEDO White Paper 19

 Having fitted the kinetic parameters, ‘what if?’ scenarios for your process can be simulated.
Unlocking Essential Process Knowledge

In Simulator you can easily alter reaction

parameters to create ‘what if?’ scenarios.

Using your fitted kinetic parameters, your process can be optimized to desired targets based on the variation of
selected parameters.

Set your optimization objectives…

…and watch the
automated optimization.

…followed by your
parameter ranges…

With fitted kinetic parameters, you can simulate the behavior of the reaction over a wide range of process conditions
using the Full Factorial function. The outputs can be displayed as a Contour Plot to understand the sensitivity of your
process to variation in reaction parameters.

20 METTLER TOLEDO White Paper

 Moving the cursor over the
contour plot allows immediate
access to the predicted outputs.

Reaction Lab enables development and testing of mechanistic hypotheses by fitting experimental data to models
with less physical experimentation required. Hypotheses can be revised directly in the software until a fit is found,
and the software can be used to design optimized processes that reduce costs and waste.

Learn more about Reaction Lab modeling software at mt.com/Reaction-Lab.

METTLER TOLEDO White Paper 21

 Want to Learn More?
Unlocking Essential Process Knowledge

Hear From Learn From Our Contact

Industry Experts Extensive Library Our Experts

Visit our webinar collection and Sign-up for Dynochem Resources, Our team of technical and applica-
listen to your peers present case our comprehensive library of models tion consultants are ready to discuss
studies demonstrating how they are and user knowledge, to learn from your application and how you can
leveraging the combined power of our experts and the community, and take steps to start accelerating your
data-rich experiments and kinetic build your expertise. development processes.
models to optimize chemical
development.
mt.com/ac-contactus
mt.com/ac-ModelingCollection dcresources.scale-up.com

METTLER TOLEDO Group

Automated Reactors and In-Situ Analysis
www.mt.com/AutoChem
For more information

Subject to technical changes.

© 10/2024 METTLER TOLEDO. All rights reserved.