Minimizing E-factor in the continuous-flow synthesis of diazepam and atropine

Article abstract of DOI:10.1016/j.bmc.2017.02.002

Minimizing the waste stream associated with the synthesis of active pharmaceutical ingredients (APIs) and commodity chemicals is of high interest within the chemical industry from an economic and environmental perspective. In exploring solutions to this area, we herein report a highly optimized and environmentally conscious continuous-flow synthesis of two APIs identified as essential medicines by the World Health Organization, namely diazepam and atropine. Notably, these approaches significantly reduced the E-factor of previously published routes through the combination of continuous-flow chemistry techniques, computational calculations and solvent minimization. The E-factor associated with the synthesis of atropine was reduced by 94-fold (about two orders of magnitude), from 2245 to 24, while the E-factor for the synthesis of diazepam was reduced by 4-fold, from 36 to 9.

Full text of DOI:10.1016/j.bmc.2017.02.002

Accepted Manuscript
Minimizing E-factor in the continuous-flow synthesis of diazepam and atropine
Anne-Catherine Bédard, Ashley R. Longstreet, Joshua Britton, Yuran Wang,
Hideki Moriguchi, Robert W. Hicklin, William H. Green, Timothy F. Jamison
PII:
S0968-0896(17)30220-1
DOI:
http://dx.doi.org/10.1016/j.bmc.2017.02.002
BMC 13529
Reference:
Bioorganic & Medicinal Chemistry
To appear in:
Received Date:
Revised Date:
Accepted Date:
14 December 2016
27 January 2017
1 February 2017
Please cite this article as: Bédard, A-C., Longstreet, A.R., Britton, J., Wang, Y., Moriguchi, H., Hicklin, R.W.,
Green, W.H., Jamison, T.F., Minimizing E-factor in the continuous-flow synthesis of diazepam and atropine,
Bioorganic & Medicinal Chemistry (2017), doi: http://dx.doi.org/10.1016/j.bmc.2017.02.002
This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers
we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting proof before it is published in its final form. Please note that during the production process
errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

Graphical Abstract
Minimizing E-factor in the continuous-flow
synthesis of diazepam and atropine
Leave this area blank for abstract info.
Anne-Catherine Bédarda, Ashley R. Longstreeta, Joshua Brittona, Yuran Wangb, Hideki Moriguchia, Robert W.
Hicklina, William H. Greenb and Timothy F. Jamisona*
a Department of Chemistry, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge,
Massachusetts 02139, USA
b Department of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue,
Cambridge, Massachusetts 02139, USA

Bioorganic & Medicinal Chemistry
journal homepage: www.elsevier.com
Minimizing E-factor in the continuous-flow synthesis of diazepam and atropine
Anne-Catherine Bédarda, Ashley R. Longstreeta, Joshua Brittona, Yuran Wangb, Hideki Moriguchia, Robert
W. Hicklina, William H. Greenb, and Timothy F. Jamisona,*
a Department of Chemistry, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA
b Department of Chemical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Avenue, Cambridge, Massachusetts 02139, USA
ARTICLE INFO
ABSTRACT
Article history:
Received
Received in revised form
Accepted
Minimizing the waste stream associated with the synthesis of active pharmaceutical ingredients
(APIs) and commodity chemicals is of high interest within the chemical industry from an
economic and environmental perspective. In exploring solutions to this area, we herein report a
highly optimized and environmentally conscious continuous-flow synthesis of two APIs
identified as essential medicines by the World Health Organization, namely diazepam and
atropine. Notably, these approaches significantly reduced the E-factor of previously published
routes through the combination of continuous-flow chemistry techniques, computational
calculations and solvent minimization. The E-factor associated with the synthesis of atropine
was reduced by 94-fold (about two orders of magnitude), from 2245 to 24, while the E-factor for
the synthesis of diazepam was reduced by 4-fold, from 36 to 9.
Available online
Keywords:
Continuous-flow chemistry
E-factor
Diazepam
2009 Elsevier Ltd. All rights reserved.
Atropine
Green chemistry
1. Introduction
the target molecule while driving towards sustainable
approaches.9,10 In facilitating both of these requirements,
A growing interest in minimizing waste generation associated
with chemical synthesis drives chemists towards more
sustainable alternatives to traditional synthesis.1,2 Various mass-
based metrics such as E-factor,3 process mass intensity (PMI),4
and atom economy5,6 have been used to define efficiencies of
chemical manufacturing processes.7,8 The E-factor was proposed
by Sheldon and is calculated for a given reaction or for a
synthetic route. It is defined as the ratio of the mass of waste
produced (excluding water) to that of product obtained (Equation
1). The number estimates the efficiency of a chemical process,
i.e. higher E-factors indicate larger amounts of waste and a less
ideal process. Many factors contribute to the increase in E-factor
in synthetic processes: (i) low yield of the desired product, (ii)
excess reagents, (iii) long, multi-step reaction sequences and (iv)
extraction and purification solvents. Usually, the value tends to
increase with the molecular complexity of the product. This trend
becomes evident when considering the increasing average E-
factor from the bulk chemical (<1 - 5), to fine chemical (5 - 50),
to pharmaceutical industries (25 - >100).3
continuous-flow chemistry has gained momentum given its
numerous advantages over traditional batch processes.11 We
propose that mindfully using flow technologies could facilitate
syntheses while significantly reducing waste production. The
benefits of flow chemistry include (i) precise control of reaction
conditions due to efficient heat and mass transfers, (ii) high
reaction reproducibility, (iii) the possibility of system
automation, (iv) safer handling of hazardous reagents along with
(v) a decreased reactor footprint.12–18
These attractive properties can minimize waste generation and
correlate with reaction efficiency.12–18 In this respect, utilizing
continuous-flow to mediate reaction telescoping can avoid
wasteful isolation and purification sequences. Taking advantage
of the rapid heat transfer present in flow systems allows for
higher reaction concentrations; less solvent is required as a heat-
transfer medium. Neat solids can also be used in continuous-flow
systems at temperatures above their melting points, removing
need for solvent.19 As reaction scale increases during the drug
design process, continuous-flow enables a straightforward and
linear scale-up with minimal re-optimization compared to
traditional batch processes. A combination of the above factors
results in a reduction in environmental impact.12
Equation (1): E(nvironmental)-factor = kg waste/kg product
The structural complexity of many active pharmaceutical
ingredients (APIs) typically implies an elaborate synthesis.2
Although in recent years continuous-flow chemistry has emerged
as a powerful tool for streamlining multi-step syntheses,13,20–28
little focus has been put on targeting waste reduction. With this in
mind, we aimed to synthesize APIs diazepam 1 and atropine 2
Numerous industries are concerned with the environmental
impacts of their operations, and yet, the poorest E-factors in the
chemical sector are found in pharmaceuticals.3 Fortuitously, the
pharmaceutical industry is continuously adapting and evolving to
improve the time between hit-to-lead discovery and production of

Scheme 1. Previously reported continuous-flow syntheses of diazepam 1 (A) and atropine 2 (B).
with dramatically lower E-factor by taking advantage of the
beneficial properties of continuous-flow chemistry. In
considering that the pharmaceutical industry produces chemicals
with E-factors ranging from 25 to more than 100, we aimed to
attain an E-factor less than 25 for both APIs.
examined to pursue this challenging target35–37, the majority of
commercial atropine is still primarily extracted from plants.32,33
Our group recently demonstrated the continuous-flow
syntheses of diazepam 1 and atropine 2.19,28 These prior syntheses
gave E-factor values of 36 and 2245 for diazepam 1 and atropine
2 respectively (Scheme 1). Herein, we report our efforts to
decrease the E-factor below 25 for each synthesis by minimizing
solvent usage, improving synthetic route to minimize byproduct
formation, and employing computational analysis to improve the
purification of the target molecule.
Diazepam 1 and atropine 2 are two APIs identified by the
World Health Organization (WHO) as essential medicines.19
Since the discovery of 1,4-benzodiazepine derivatives as central
nervous system depressants, considerable research has been
carried out on compounds featuring this interesting seven-
membered ring system.29 Diazepam 1, commercially known as
Valium, is used to treat anxiety and epilepsy.30 First synthesized
by Leo Sternbach at Hoffman-La Roche in 1959, it entered the
market in 1963 and became the top selling pharmaceutical in the
United States between 1968 and 1982.31 Atropine 2 (D/L-
hyoscyamine) is a naturally occurring alkaloid of the tropane
2. Results and Discussion
2.1. The synthesis of diazepam 1
The previously reported continuous-flow synthesis of diazepam
began with an amidation reaction between bromoacetyl chloride
4 and 5-chloro-2-(methylamino)benzophenone 3 in N-methyl-2-
pyrrolidone (NMP) to produce a mixture of amides 5 and 6
(Scheme 1A).19 Next, a solution of ammonia in MeOH / H2O was
mixed with the reaction stream to cyclize amides 5 and 6 into
diazepam 1. The crude mixture then underwent two subsequent
in-line extractions. The first extraction removed water-soluble
impurities through the addition of an aqueous sodium chloride
family found in several solanaceous plants.32,33 It is
a
competitive, non-selective antagonist of muscarinic acetylcholine
(M2) receptors. It serves as a preoperative and sedation drug for
short-term medical procedures.34 This molecule was first
synthesized in 1879 by Landenburg via a Fischer Speier
esterification of tropine with tropic acid in the presence of
hydrochloric acid.35 Although several syntheses have been

(NaCl, 20 wt. %) and ethyl acetate (EtOAc). With diazepam 1
residing in the organic phase, separation from the aqueous phase
was achieved by a gravitational phase-separator. A 78% yield
was obtained after the first extraction. The organic stream was
then passed through a charcoal column and mixed with a stream
of aqueous HCl (4 M) to protonate diazepam 1 and enable its
extraction into the aqueous phase. This second in-line extraction
removed organic soluble impurities from the product stream.
Next, protonated diazepam 1 was separated from the organic
stream through an additional gravitational phase-separator. The
product was isolated in a pure form after a continuous
recrystallization process using downstream neutralization with
aqueous ammonium hydroxide.
ClCH2COCl 14 addition also resulted in lower yields (Entries 4-
7). The E-factor for this step was improved slightly by using neat
ClCH2COCl 14 instead of a 2.7 M solution of BrCH2COCl 4 in
NMP. Furthermore, this approach led to a four-fold decrease of
the reaction time (from 9.3 min down to 2.4 min).
Table 1. Optimization of amidation.
The calculated E-factor up to the first work-up is 36 for this
process (Scheme 1A); our target was to decrease it below 25.
Solvent is often the primary contributor to E-factor3, and in this
instance, extraction with EtOAc contributes to two thirds of the
waste produced. In order to eliminate the need for a separate
extraction solvent, we investigated a solvent that could be used
for both reaction and extraction purposes. We chose 2-
methyltetrahydrofuran (2-MeTHF) for several factors: (i) the
solubility of the starting materials and most products were within
a reasonable range to enable reaction concentrations >0.9 M, (ii)
the low density at 0.85 g/mL minimizes the mass of waste
produced, (iii) 2-MeTHF is not reactive with the reagents at high
temperatures, and (iv) 2-MeTHF is immiscible with water, i.e.
enables an in-line aqueous extraction without additional solvent.
ClCH2COCl 14
Flow rate (μL/min)
Reactor I
temperature (°C)
ClCH2COCl 14 Yield
Entry
Equiv
(%)a
1
2
3
4
5
6
7
70
80
90
80
90
15.0
15.0
15.0
13.5
13.5
13.5
12.0
1.25
1.25
1.25
1.13
1.13
1.13
1.00
79
98
99
89
92
88
83
120
120
aIsolated yields. Reaction was quenched with aqueous saturated NaHCO3
solution within the collection vial.
With the amidation reaction optimized, we then investigated
the telescoped synthesis of diazepam 1 (Table 2). When
determining a suitable ammonia source, aqueous ammonium
hydroxide was chosen to avoid addition of excess organic
solvent. In addition, an aqueous stream was necessary to prevent
clogging due the formation of ammonium chloride. Upon the
addition of ammonium hydroxide, amide 5 precipitated out of
solution. To prevent clogging at the T-mixer, ammonium
hydroxide addition was performed in a heated sonicating bath. At
After identifying 2-MeTHF as an ideal solvent, we began
screening conditions for the amidation step between a solution of
benzophenone 3 (1 M in 2-MeTHF) and neat chloroacetyl
chloride (ClCH2COCl) 14 (Table 1). To prevent clogging at the
end of the reactor upon cooling or at the BPR, dichloromethane
(DCM) was used to dilute the reaction stream. As demonstrated
in the previous flow process, the amidation proceeds even in the
absence of a base. Upon heating to 80 °C with 1.25 equiv. of
ClCH2COCl 14, amide 5 was obtained in near quantitative yields
(Table 1, Entries 2 and 3). Decreasing the temperature to 70 °C
resulted in a 20% decrease in yield (Entry 1), and decreasing the
o
60 C, no reaction occured in the sonicating bath. To ensure
efficient mixing had happened between the aqueous and organic
phases, a static mixer was installed between the ammonium
Table 2. Cyclization optimization to produce diazepam 1.
Entry NH4OH (NH3 28 – 30 wt. %) to H2O vol. ratio NH4OH Equiv. Reactor II temperature (°C) Yield (%)a
No dilution 23 90 49
No dilution 23 100
NAb
1
2
3
4
1:1
3:1
3:1
3:1
4:1
4:1
4:1
9:1
9:1
11
14
14
14
18
18
18
21
21
120
100
110
120
100
110
120
100
110
41
46
5
51
NAb
6
7
46
8
41
NAb
9
10
11
55
NAb

hydroxide addition and Reactor II.
maximize conversion (Table 3). In the previous synthesis,
tropine•HCl 7 and phenylacetyl chloride 8 were added from a
single stock solution.28 This forced the use of the tropine•HCl salt
7 to prevent unwanted reactivity prior entry into the reactor.
Using two distinct incoming streams, i.e. tropine 9 in DMF and
neat phenylacetyl chloride 8, allowed use of the tropine free base
in a 10-fold concentration increase. Using 1.05 equiv. of
phenylacetyl chloride 8 as a 2 M solution in DMF gave 80%
conversion in a 10 minutes residence time (Entry 1). The addition
of larger ratios of phenylacetyl chloride 8 increased conversion to
87% (Entry 2). Complete conversion was obtained in 10 minutes
when using neat phenylacetyl chloride 8 (Entry 3). Notably, the
residence time could be further decreased to 3.5 minutes while
maintaining complete conversion (Entry 5). This improvement of
reaction time also prevented the formation of byproduct 13.
Satisfyingly, tropine 9 was able to be used in concentrations
close to saturation and phenylacetyl chloride 8 was used neat.
This ensured high throughput while reducing waste and solvent
quantities. Additionally, the use of two streams of reagents
allowed for the tertiary amine of ester 15 to act as an internal
base for the reaction, removing the need for additional base in the
second step.
The initial screening was performed with concentrated
ammonium hydroxide solution (ca. 28-30 wt. % based in NH3)
with a reactor temperature of 90 °C, which resulted in a 49%
yield (Table 2, Entry 1). When the reactor temperature was
increased to 100 °C, violent gas evolution was observed in the
reactor and no product was formed (Entry 2). Because neither 2-
MeTHF or water alone would result in gas evolution when
passed through the heated reactor, we hypothesized the gas
present was gaseous ammonia. To prevent ammonia gas
evolution, the backpressure was first increased. However,
installing backpressures >100 psi resulted in pump failure. We
therefore screened different dilutions of the ammonium
hydroxide solution. By diluting the concentrated NH4OH solution
in a 1:1 vol. ratio with water, Reactor II could operate at 120 °C
without gas evolution (Entry 3). However, a concomitant drop in
yield to 41% prompted screening both ammonium hydroxide
dilutions and reactor temperatures. When the solution was diluted
to a 3:1 or 4:1 volume ratio in water, the reactor could be heated
to 110 °C without gas evolution (Entries 4-9) with only minimal
changes in yield observed. Diluting the solution to a 9:1 vol. ratio
in water and passing the resulting solution through a reactor
heated to 100 °C provided a 55% yield (Entry 10). Further
heating the reactor to 110 °C resulted in detrimental gas
evolution (Entry 11). With a 55% yield of diazepam 1 in hand,
our target E-factor was reached at 9. Furthermore, experiments
demonstrated that the static mixer before Reactor II could be
removed since a negligible change in yield from 55% to 51% was
obtained (Table 2, Entry 12).
Table 3. Esterification optimization.
Improving an already efficient synthesis was not trivial, but
overall, the E-factor for the synthesis of diazepam 1 was
decreased by 4-fold (from 36 down to 9) and now resides in the
range of fine chemicals processes. A key point contributing to
this success was the simplified setup achievable upon using 2-
MeTHF as both the reaction and extraction solvent.
Cl(CO)CH2Ph 8 Cl(CO)CH2Ph 8
tR
(min)
Entry
Conversion (%)a
equiv.
conc. (M)
1b
2b
3
4
5
1.05
1.10
1.00
1.00
1.00
2.0c
2.0c
7.6d
7.6d
7.6d
10
10
10
7.0
3.5
80
87
99
99
99
2.1. The synthesis of atropine 2
Our previous synthesis of atropine 2 (Scheme 1B)28
reported in 2015 was unfortunately met with the formation of
several byproducts. Although initial esterification of the
aGCMS conversion. b40 psi BPR. cSolution in DMF. dNeat.
Next, the challenging optimization of the aldol addition to
produce atropine 2 was examined (Table 4). Previously, the use
of super-stoichiometric amounts of sodium hydroxide (4.4
equiv.) favored E1CB elimination to form apoatropine 11 and
lead to the cleavage of the tropine ester bond to yield tropine 9.28
Byproduct 12 was a result of a double addition of formaldehyde,
again favored by the use of super-stoichiometric amount of
formaldehyde (3 equiv.). It was found that alterations made to the
esterification reaction (increasing concentration, modifying
stoichiometry, and switching from tropine•HCl 7 to tropine free
base 10) significantly affected the tandem sequence. The DMF
solution exiting the first reactor was oversaturated and the tubing
required heating (>50 oC) to prevent crystallization of the mixture.
This highlights another advantage of continuous-flow which is
the possibility of pumping a solid above its melting point. Upon
addition of NaOH and formaldehyde, formation of a slug-flow
type regime was observed as the incoming stream is only
sparingly miscible with the reagents needed for the second step.
A static mixer was added to improve the mixing before entering
the second reactor.
tropine•HCl salt
7
with phenylacetyl chloride
8
(Cl(CO)CH2Ph) gave the corresponding ester 10 in complete
conversion, the subsequent aldol condensation using aqueous
formaldehyde and aqueous sodium hydroxide (3 M NaOH)
was challenging. Atropine 2 is very sensitive to elimination
affording the thermodynamically more stable apoatropine 11.
Furthermore, the tropine ester bond is prone to saponification
in basic conditions yielding tropine 9. Additionally, the
double aldol adduct 12 is observed concomitantly with DMF
degradation adduct 13. Since these byproducts are structurally
similar, an extensive set of in-line extractions were needed to
obtain the desired product in high purity. The mass of DCM
used for the three extractions contributed up to 63% of the
waste generated and gave rise to an E-factor of 2245. The
other main contributor to the high E-factor for this synthesis is
the low overall yield of atropine 2 obtained (8%). In order to
decrease the E-factor for the synthesis of atropine 2, several
aspects were investigated herein: (i) improvement of the
esterification transformation, (ii) minimization of byproduct
formation in the aldol transformation, and (iii) design of a
new extraction procedure avoiding the use of DCM.
First, the esterification was optimized to increase
concentration, minimize equivalents, and reaction time, and

Table 4. Aldol optimization to produce atropine 2.
Entry
NaOH Conc (M); Equiv
H2CO Equiv
Reactor II temperature (°C)
Reactor II tR (min)a
Conversion (%)b,c
Ratio 11:2:10b
0:0:1
1
2
3; 3
3; 3
6
6
6
6
6
6
6
6
6
3
23
8
8
0
30
50
0.14:1:2.7
0.8:1:1.6
0.88:1:2
3
3:3
50
24
8
53
4
3; 3
100
150
100
100
100
100
100
49
5
3; 3
8
67
78(44e)
2.7:1:1.8
0.64:1:0.45
0.65:1:0.8
0.8:1:0.6
0.54:1:0.85
0.5:1:2.3
6
3;1.2
3;1.2
1;1.2
1;1.2
1;1.2
24
8
7
67
8
8
75
64d
9
8
10
8
40
aReactor size was adjusted to match the desired residence time. bDetermined by 1H NMR of the crude mixture. cConversion of tropine ester 10 to atropine 2
and apoatropine 11. dStatic mixer was removed. eNMR yield using trimethoxybenzene as an internal standard.
U
nder these modified conditions, no conversion of the tropine
In SPARC, coefficients are calculated from algorithms analyzing
quantitative structure-activity relationships through linear free-
energy relationships and quantum effects based around
perturbation molecular orbital methods. A variety of molecular
descriptors are harnessed to calculate solute-solvent interaction
based upon dispersion, induction, dipole-dipole and hydrogen
bonding interactions. The logD values generated are calculated
with respect to water at standard temperature and pressure.
ester 10 to the desired product 2 was observed at room
temperature (Entry 1) when three equiv. of 3 M NaOH were
used. Furthermore, the byproduct arising from double
hydroxymethylation 12 was not observed. Increasing the
temperature to 50 oC gave a 30% conversion for a residence time
of eight minutes (Entry 2). Further increasing the temperature to
o
100 C or lengthening the reaction time to 24 minutes increased
both the conversion to 49% and 53%, respectively (Entries 3 and
4). A further increase in conversion was observed at 150 C, but
The calculated values suggested that DCM was not an ideal
extraction solvent in the process since atropine 2 is soluble in it at
all pH values. However, greener solvents could be used to
achieve good separation (Figure 1A-D). The computational
models indicate that diethyl ether, toluene, butyl acetate, 2-
o
higher temperatures led to augmented elimination to 11 (Entry 5.)
Decreasing the amount of base to 1.2 equiv. (i.e. 0.2 equiv. for
the aldol reaction and 1 equiv. for the deprotonation of 15) at 100
oC afforded 78% conversion with a 0.64 to 1 ratio of 11 to 2
(Entry 6). Notably, the reaction time could be decreased to 8
minutes when using a 1 M solution of NaOH while maintaining a
conversion of 75% (Entry 8). Achievement of this lower reaction
time is likely due to improved mixing between the slugs enabled
by the additional water in the reactor. The static mixer was still
needed under these conditions as about 10% conversion was lost
upon its removal (Entry 9). Attempts to decrease the amount of
formaldehyde resulted in reduced conversion (Entry 10).
t
MeTHF, N-butyl ether and butyl methyl ester have appropriate
extraction properties to isolate atropine 2 from the mixture. At
pH values below 7, atropine 2 and byproduct 9 would remain in
the aqueous phase while the byproducts 10 and 11 would
predominantly transfer to the organic phase. Next, simply
adjusting the pH of the solution to >9 would drive atropine 2 into
the organic phase while byproduct 9 stays in the aqueous layer.
This pH-controlled extraction was tested with diethyl ether, butyl
t
acetate and butyl methyl ester under typical batch techniques
(Supplementary Information). Diethyl ether was optimal with
atropine 2 remaining in the aqueous phase at pH 6.5 and then
transferring quantitatively into the organic phase at pH 10.
The two-step sequence to obtain atropine 2 was improved and
a yield of 44% was obtained under optimal condition (Table 4,
Entry 6). A better control of the reaction was achieved and half of
the undesired byproducts were eliminated from the product
distribution. The E-factor obtained before the in-line extraction is
seven, a significant improvement compared to 123 obtained at
the same point in the previous synthesis.
Next, this extraction process was translated into continuous-
flow (Figure 2E). In analyzing the performance of this process, a
1:1 mixture of atropine 2 and apoatropine 11 were mixed in
buffer (pH 6.5, NH4Cl, 0.42 M) and flowed through a Zaiput
liquid-liquid separator module with diethyl ether as the organic
phase. In agreement with both the computational calculations and
our batch model, quantitative levels of the byproduct
(apoatropine 11) were extracted into diethyl ether, leaving only
atropine 2 in the aqueous phase. Although atropine 2 could be
further abstracted into organic solvents at >pH 9 (Supporting
Information), it can also be crystalized from the aqueous phase in
the future downstream processes.40 The decision to isolate
Next, the in-line purification of atropine 2 was explored to
identify an optimal extraction solvent while minimizing the
quantity of solvent required. We envisaged that controlling the
pH of the product stream would allow atropine to be separated
from its byproducts based upon differences in pKa. To expedite
this process, the distribution coefficients of atropine 2 and its
byproducts in various organic solvents were calculated from pH
1-14 using SPARC computational modeling (Figure 1A-D).38,39

Figure 1. Computational data for the effective separation of atropine 2 from byproducts 9, 10, and 11 and the model continuous-flow
purification platform. A) The LogD values for atropine 2 in a range of solvents. As shown here, DCM is less effective in retaining
atropine 2 under acidic conditions. B) The LogD values for the atropine ester byproduct 10 in a range of solvents. This byproduct is
soluble at all pH values in DCM. C) The LogD values for apotropine 11 in a range of solvents. This species is soluble in DCM at all pH
values. D) The LogD values for tropine 9 in a range of solvents. E) The data generated in these computational plots was used as a
guideline to design an effective continuous-flow separation. Batch related separation experiments to derive ether as the most effective
solvent are available in the Supplementary Information.
atropine in the aqueous phase further lowered the E-factor by
removing the need for additional organic solvent while
decreasing reactor footprint.
This purification model was then tested on the optimized flow
system (Table 5). Clogging was observed when pH 6.5 buffer
was added to the reaction stream. A cross mixer was used in
order to add the diethyl ether and the buffer at once and prevent
precipitation. When a 1:1 ratio of pH 6.5 buffer to incoming
reaction stream was used, the desired extraction pH was not
reached and only traces of atropine 2 were extracted in the
aqueous layer (Entry 1). Increasing the flow rate of buffer
allowed the system to reach the desired extraction pH, i.e. 6.5
(Entry 2). Disappointingly, the selectivity for the exclusive
extraction of atropine 2 was lost. It is presumed that the residual
DMF coming from the esterification reaction affected the product
distribution during the extraction. We then investigated the
addition of a packed-bed to aid with mixing. A 6 cm packed-bed
containing stainless steel (SS) beads was used with no
Table 5. Optimization of in-line atropine 2 extraction.
Aqueous quench;
mL/min
Solvent;
mL/min
Packed-bed content;
length (cm)
pH of aqueous
stream
Aqueous stream
contenta
Entry
Organic stream contentb
1
2
pH 6.5 bufferc, 0.3c
pH 6.5 bufferc, 0.4
pH 6.5 bufferc, 0.4
pH 6.5 bufferc, 0.4
pH 6.5 bufferc, 0.4
water, 0.3
Et2O, 0.6
Et2O, 0.6
NA
NA
9
traces 2
2+10+11
2+10+11
2+10+11
2+10+11
2+10+11
2+10+11
2e (22%f)
2+10+11
2+10+11
6.5
6.5
6.5
6.5
11
3
Et2O, 0.6
SS beads; 6
Sand; 23
Sand; 6 (x2)
Sand; 6
2+10+11
4
5d
Et2O, 0.6
2+10+11
Et2O, 0.3 (x2)
Et2O, 0.15
DCM, 0.15
PhMe, 0.15
2+10+11
6
2+10+11
7
water, 0.3
Sand; 6
11
2+10+11
2+10+11 (1:0.95:0.72g)
8
water, 0.3
Sand; 6
11
a Determined by 1H NMR after basification to pH 10 adjustment and DCM extraction. b Determined by 1H NMR. c NH4Cl buffer (0.42 M) d1:1 ratio of
6.5 buffer to incoming reaction stream (295 uL/min). Reaction stream was met with 6.5 buffer stream and 1st Et2O stream (0.3mL/min), mixed through
a 6 cm packed-bed, extracted once. The resulting aqueous stream was met with 2nd Et2O stream (0.3 mL/min), mixed in a second 6 cm packed-bed and
extracted again. e Contains 4% of 11 and 6% of 10 by 1H NMR. f Yield based on 1H NMR using trimethoxybenzene as an internal standard.g Ratio of

improvement in extraction efficiency (Entry 3). A longer (23 cm)
packed-bed containing sand was used without any improvement
in extraction efficiency but did lead to a smaller pressure drop in
the system compared to the SS bead column (Entry 4).
Performing two extractions with Et2O likewise resulted in an
undesired mixture of products. Re-examining the data in Figure 1
indicated that atropine 2 has a greater affinity for water at
elevated pH compared to the other byproducts such as
apoatropine 11 and the tropine ester 10, with the exception of
tropine 9. This observation is also consistent with experimental
data (Table 5, Entry 1). Initially, an extraction at basic pH was
not considered due to the possible elimination of atropine 2 to
apoatropine 11 or ester cleavage to yield tropine 9. The similar
solubility of atropine 2 and tropine 9 at elevated pH was also a
concern for using this approach. Fortunately, our optimized
conditions for the synthesis of atropine 2 did not yield any
residual tropine 9 and we were hence encouraged to explore
extraction under basic condition (pH >8). Although diethyl ether
and DCM did not yield a clean extraction (Entries 6 and 7), no
additional apoatropine 11 was formed in the process.
Gratifyingly, toluene (PhMe) was able to selectively extract
atropine 2 from the mixture (Entry 8). This finding corroborates
the computational data (Figure 1A) that indicated a smaller logD
for atropine 2 in PhMe at pH 11.
0.20 μm, and stainless steel 1/16″ female nut from
1
SWAGELOCK®. H NMR and 13C NMR spectra were recorded
at ambient temperature using CDCl3 (7.27 ppm) or d6-DMSO
(2.50 ppm), unless otherwise indicated on a Varian 500 MHz
spectrometer or Bruker 400 MHz spectrometer.
4.2 Synthetic procedures
Diazepam 1:
A
25 mL solution of 5-chloro-2-
(methylamino)benzophenone (6.14 g, 1 M) and dodecane (0.57
mL, 0.1 M) in 2-MeTHF was loaded into a SS syringe and
pumped via syringe pump into the system at 0.15 mL/min and
neat chloroacetyl chloride was loaded into a SS syringe and
pumped via syringe pump at 0.015 mL/min. The streams met in a
T-mixer (0.02” ID) before entering a reactor made from 0.04” ID
o
PFA tubing (0.4 mL volume) preheated to 90 C in an oil bath.
Upon exiting the reactor, the stream was met with aqueous
NH4OH solution (28-30 wt % diluted in water in a 9:1 volumetric
ratio) pumped at 0.234 mL/min via syring pump in a sideways T-
mixer (0.04” ID) that was sitting in a sonication bath preheated to
o
60 C. Upon exiting the sonicating bath, the stream entered the
second reactor made of 0.04” ID PFA tubing (10.2 mL volume)
that was preheated to 100 oC in an oil bath. An in-line separation
was then performed using a membrane separator containing a 1.0
μm pore PTFE microfiltration membrane. Two BPRs set to 100
psi were installed at the end of the reactor on each side of the
membrane separator. The entire system was equilibrated for 1 h,
and product collection lasted for 1 h. The solution collected
contained diazepam with a 49% calibrated yield determined by
GC analysis with dodecane as the internal standard. This solution
was then passed through a packed-bed of activated charcoal at
0.15 mL/min, then a sample (0.90 mL) was collected and
diazepam 1 was isolated by automated flash chromatography (Rf
= 0.19 in 30% EtOAc/hexane) as a colorless solid (0.103 g, 40%
Overall, the full synthesis of atropine 2, including the
extraction, gave a 22% yield of the desired product. This
transposes to nearly a 3-fold increase in yield and a significantly
improved E-factor of 24, a 94-fold difference from the previous
route.
3. Conclusion
The expedited synthesis of APIs has been aided by the
development of continuous-flow approaches. However, the
environmental consequences of syntheses are rarely considered.
We presented efficient E-factor optimization strategies for the
synthesis of two essential APIs: atropine and diazepam. Notably,
our approaches significantly reduced the waste generated in both
routes through the combination of continuous-flow chemistry
techniques, computational calculations and solvent minimization.
Diazepam waste was decrease by four-fold and now resides in
the range of fine chemicals processes. Atropine displays an
impressive two orders of magnitude improvement in term of
sustainability. We believe the continuous-flow technology and
rational strategies discussed herein to reduce the E-factor pave
the way for further APIs syntheses that are more environmentally
conscious.
1
yield). H NMR and 13C NMR in CDCl3 are in accordance with
reported literature values.14
Atropine 2: The solution of tropine (2 M in DMF) was loaded
in two stainless steel (SS) syringe. Neat phenylacetyl chloride
was loaded into a dry 5 mL luer-lock glass syringe. The solutions
where pumped (using a Harvard Apparatus Syringe Pump) at
69.2 μL/min and 18.3 μL/min respectively. The solutions met in
a T-mixer (0.02” id) at room temperature and were then warmed
o
to 100 C in a 300 μL reactor (87.5 μL/min total flow rate = 3.5
min residence time). Importantly, the stream exiting the T-mixer
had to be rapidly warmed to 100 oC, otherwise clogging occurred.
The stream exiting the first reactor was then connected to a T-
mixer (0.02” id) and met a stream of 3 M NaOH pumped by an
SS HPLC pump (56 μL/min, 1.2 equiv.). A 15 cm (0.02” id)
segment of tubing was placed before the 2nd T-mixer where
formaldehyde (37% aq, 65.4 μL/min, 6 equiv.) was added using a
Harvard Apparatus Syringe Pump and two 8 mL SS syringe. The
reaction mixture (209 μL/min total flow rate) was then reacted at
4. Experimental
4.1 Material and methods
Unless otherwise indicated, all commercially available
reagents and solvents were used directly from the supplier
without further purification. Membrane liquid-liquid separators
and backpressure regulators (BPRs) were purchased from Zaiput
Flow Technologies, and the PTFE microfiltration membranes
were bought from Pall Zefluor with 0.5 or 1.0 μm pore size. The
reactors were constructed from high-purity perfluoroalkoxy
(PFA) tubing with 1/16” OD and 0.03″ ID and PEEK fittings
purchased from IDEX Health & Science Technologies, unless
otherwise noted. Harvard Apparatus PhD Ultra syringe pumps
were used to pump reagents and solutions from 8-mL high-
pressure stainless steel syringe with 1/16” SWAGELOCK® from
Harvard Apparatus unless otherwise noted. A 5 mL glass syringe
was used with phenylacetyl chloride due to incompatibility with
stainless steel. The packed-bed scavenger was assembled from a
stainless steel tube (1/4” OD, 3/16” ID), Stainless Steel (SS) frits
o
100 C for 24 min in a 5 mL reactor (0.03” id) that was capped
with a 100 psi BPR. After exiting the BPR, the reaction stream
was mixed using a cross-mixer (0.02” id) both water (0.3
mL/min, HPLC pump) and PhMe (0.15 mL/min, pumped using a
MilliGat M6 pump purchased from Valco Intruments Co. Inc.).
The resulting slug-flow mixture (0.659 mL/min) was passed
through a 6 cm SS sand-filled packed-bed for mixing purposes.
An in-line extraction using a Zaiput liquid-liquid extractor
(membrane = 0.5 μm) was achieved and both the aqueous and
organic streams were collected in separate scintillation vials. The
system was equilibrated for 1h, then the products were collected
for 20 min. The organic stream was collected in an empty vial
but the aqueous stream (measured at pH 11) was collected in a
vial containing DCM to extract the product from water and

M.; Snead, D. R.; Stelzer, T.; Weeranoppanant, N.; Wong, S. Y.;
Zhang, P. Science 2016, 352 (6281), 61.
Lévesque, F.; Seeberger, P. H. Angew. Chem. Int. Ed. 2012, 51 (7),
1706.
Hopkin, M. D.; Baxendale, I. R.; Ley, S. V. Org. Biomol. Chem.
2013, 11 (11), 1822.
Correia, C. A.; Gilmore, K.; McQuade, D. T.; Seeberger, P. H.
Angew. Chem. Int. Ed. 2015, 54 (16), 4945.
Viviano, M.; Glasnov, T. N.; Reichart, B.; Tekautz, G.; Kappe, C.
O. Org. Process Res. Dev. 2011, 15 (4), 858.
Zhang, P.; Russell, M. G.; Jamison, T. F. Org. Process Res. Dev.
2014, 18 (11), 1567.
prevent elimination/degradation. The organic stream was
concentrated in vacuo. A mixture of 2:10:11 was obtained (1H
NMR ratio 1:0.95:0.72) 332 mg total mass, 22% yield of 2. The
aqueous stream was extracted with DCM, concentrated in vacuo.
(20)
(21)
(22)
(23)
(24)
(25)
(26)
(27)
Atropine 2 was obtained in 176 mg, 22% yield. 1H NMR and 13
C
NMR in CDCl3 are in accordance with reported literature
values.41
Acknowledgments
Ghislieri, D.; Gilmore, K.; Seeberger, P. H. Angew. Chem. Int. Ed.
2015, 54 (2), 678.
This work was supported by the Defense Advanced Research
Project Agency (DARPA), Army Research Office under contract
number W911NF-16-0023. We thank R. Beingessner for her
advice and support. ACB thanks NSERC for a postdoctoral
fellowship. HM thanks ONO pharmaceutical Co., Ltd. for their
support.
Murray, P. R. D.; Browne, D. L.; Pastre, J. C.; Butters, C.; Guthrie,
D.; Ley, S. V. Org. Process Res. Dev. 2013, 17 (9), 1192.
Heider, P. L.; Born, S. C.; Basak, S.; Benyahia, B.; Lakerveld, R.;
Zhang, H.; Hogan, R.; Buchbinder, L.; Wolfe, A.; Mascia, S.;
Evans, J. M. B.; Jamison, T. F.; Jensen, K. F. Org. Process Res.
Dev. 2014, 18 (3), 402.
(28)
Dai, C.; Snead, D. R.; Zhang, P.; Jamison, T. F. J. Flow Chem.
2015, 5 (3), 133.
References and notes
(29)
(30)
Sternbach, L. H. Angew. Chem. Int. Ed. Engl. 1971, 10 (1), 34.
Calcaterra, N. E.; Barrow, J. C. ACS Chem. Neurosci. 2014, 5 (4),
253.
Sternbach, L. H. J. Med. Chem. 1979, 22 (1), 1.
Dimitrov, K.; Metcheva, D.; Boyadzhiev, L. Sep. Purif. Technol.
2005, 46 (1–2), 41.
(1)
(2)
Beach, E. S.; Cui, Z.; Anastas, P. T. Energy Env. Sci 2009, 2 (10),
1038.
Roschangar, F.; Sheldon, R. A.; Senanayake, C. H. Green Chem.
2015, 17 (2), 752.
Sheldon, R. A. Green Chem 2007, 9 (12), 1273.
Jimenez-Gonzalez, C.; Ponder, C. S.; Broxterman, Q. B.; Manley, J.
B. Org. Process Res. Dev. 2011, 15 (4), 912.
Trost, B. Science 1991, 254 (5037), 1471.
Vaxelaire, C.; Winter, P.; Christmann, M. Angew. Chem. Int. Ed.
2011, 50 (16), 3605.
Dach, R.; Song, J. J.; Roschangar, F.; Samstag, W.; Senanayake, C.
H. Org. Process Res. Dev. 2012, 16 (11), 1697.
Constable, D. J. C.; Curzons, A. D.; Cunningham, V. L. Green
Chem. 2002, 4 (6), 521.
(31)
(32)
(3)
(4)
(33)
Gadzikowska, M.; Grynkiewicz, G. Acta Pol. Pharm. 2002, 59 (2),
149.
(5)
(6)
(34)
(35)
(36)
(37)
Mucklow, J. C. Br. J. Clin. Pharmacol. 2000, 49 (6), 613.
Ladenburg, A. Berichte Dtsch. Chem. Ges. 1879, 12 (1), 941.
Robinson, R. J. Chem. Soc. Trans. 1917, 111 (0), 762.
Gore, V.; Joshi, R.; Jadhav, M.; Bhandari, S. US Patent
2014/102829 Al, 2014.
Hilal, S. H.; Karickhoff, S. W.; Carreira, L. A. Quant. Struct.-Act.
Relatsh. 1995, 14 (4), 348.
Hilal, S. H.; Karickhoff, S. W.; Carreira, L. A. QSAR Comb. Sci.
2004, 23 (9), 709.
Trissel, L. A.; American Pharmacists Association. Trissel’s stability
(7)
(8)
(38)
(39)
(40)
(9)
Sharma, S. K. Green chemistry for environmental sustainability;
CRC Press: Boca Raton, 2010.
Kockmann, N.; Gottsponer, M.; Zimmermann, B.; Roberge, D. M.
Chem. Eur. J. 2008, 14 (25), 7470.
(10)
of compounded formulations
(11)
(12)
(13)
Malet-Sanz, L.; Susanne, F. J. Med. Chem. 2012, 55 (9), 4062.
Newman, S. G.; Jensen, K. F. Green Chem. 2013, 15 (6), 1456.
Gutmann, B.; Cantillo, D.; Kappe, C. O. Angew. Chem. Int. Ed.
2015, 54 (23), 6688.
http://online.statref.com/Notes/ResolveNote.aspx?NoteID=64226&
grpalias=Alberta.
Do Pham, D. D.; Kelso, G. F.; Yang, Y.; Hearn, M. T. W. Green
Chem. 2012, 14 (4), 1189.
(41)
(14)
(15)
Pashkova, A.; Greiner, L. Chem. Ing. Tech. 2011, 83 (9), 1337.
Wegner, J.; Ceylan, S.; Kirschning, A. Chem. Commun. 2011, 47
(16), 4583.
Supplementary Material
(16)
(17)
Pastre, J. C.; Browne, D. L.; Ley, S. V. Chem. Soc. Rev. 2013, 42
(23), 8849.
Hartman, R. L.; McMullen, J. P.; Jensen, K. F. Angew. Chem. Int.
Ed. 2011, 50 (33), 7502.
Glasnov, T. N.; Kappe, C. O. Chem. Eur. J. 2011, 17 (43), 11956.
Adamo, A.; Beingessner, R. L.; Behnam, M.; Chen, J.; Jamison, T.
F.; Jensen, K. F.; Monbaliu, J.-C. M.; Myerson, A. S.; Revalor, E.
Supplementary data associated with this article can be found,
in the online version.
(18)
(19)