Why we might be misusing process mass intensity (PMI) and a methodology to apply it effectively as a discovery level metric

Article abstract of DOI:10.1039/c9gc01537j

Process mass intensity (PMI) is a key mass-based metric to evaluate the green credentials of an individual or sequence of reactions during process and chemical development. The increasing awareness to consider greenness as early as the initial discovery level, requires a set of parameters suitable to assess it at this stage of development, and guidelines to apply them correctly. This paper evaluates when and how PMI can be used in a correct manner. Different simulations for key reactions in the organic synthesis toolbox-i.e. amide bond formation and Mitsunobu reactions-illustrate that PMI can easily be misleading without due consideration of yield, concentration and molecular weight of reactants and product. A fair appraisal of the green potential of different methodologies therefore requires careful analysis of the examples and metrics data generated.

Full text of DOI:10.1039/c9gc01537j

Green Chemistry
View Article Online
View Journal | View Issue
PAPER
Why we might be misusing process mass intensity
PMI) and a methodology to apply it effectively as
a discovery level metric†
(
Cite this: Green Chem., 2020, 22,
23
1
a
b
a
Edward R. Monteith, ‡ Pieter Mampuys, ‡ Louise Summerton,
James H. Clark,
a
b
a
Bert U. W. Maes * and C. Robert McElroy
*
Process mass intensity (PMI) is a key mass-based metric to evaluate the green credentials of an individual
or sequence of reactions during process and chemical development. The increasing awareness to con-
sider greenness as early as the initial discovery level, requires a set of parameters suitable to assess it at
this stage of development, and guidelines to apply them correctly. This paper evaluates when and how
PMI can be used in a correct manner. Different simulations for key reactions in the organic synthesis
toolbox – i.e. amide bond formation and Mitsunobu reactions – illustrate that PMI can easily be mislead-
ing without due consideration of yield, concentration and molecular weight of reactants and product. A
fair appraisal of the green potential of different methodologies therefore requires careful analysis of the
examples and metrics data generated.
Received 8th May 2019,
Accepted 7th November 2019
DOI: 10.1039/c9gc01537j
rsc.li/greenchem
about the reaction efficiency, process or conditions. There
have been a considerable number of green chemistry metrics
Introduction
Over the last decade pressure on the chemical industry to developed over the past few decades, some of the most
8
change their practices and develop more environmentally common are atom economy (AE), environmental factor
9
10
friendly and sustainable technologies has drastically (E-factor) and reaction mass efficiency (RME). Other metrics
1
1
increased. While the fine chemicals industry typically pro- which appear frequently are process mass intensity (PMI),
1
2
10
duces a large amount of waste, with E-factors in the range of mass intensity (MI), carbon efficiency (CE), atom utiliz-
1
13
14
2
5 to >100 kg kg− for pharmaceuticals, product volumes are ation (AU), environmental quotient (EQ). There are many
1
5–17
low in comparison to commodity chemicals featuring lower more green measures of success in circulation,
E-factors (<1–5 kg kg ). Both sectors therefore require a focus Clark group summarized in a comprehensive list of over 60
which the
−
1
1
–3
18
on sustainability.
different metrics. Recently, Roschangar et al. made a strong
Initially the efficiency and sustainability of a process was a push to unify green mass-based metrics in industry via the
1
9–21
difficult concept to measure and quantify, which lead to the introduction of a green aspiration level (GAL).
development of green chemistry metrics.
This
4
–7
The accepted con- approach works by taking the average E-factor of active
sensus for any metric is that it must be simple, easily measur- pharmaceutical ingredients (API) at an early stage of develop-
able and clearly highlight the desired information. From a his- ment but not include solvents or water, termed simple E-factor
torical viewpoint the original metric was percentage yield; (sEF) and the average of commercial processes, capturing all
while this is good at conveying the amount of final product inputs, including process water, but not considering recycling,
1
9–21
obtained from the limiting reactant, it revealed nothing else termed complete E-factor (cEF).
These values are 42 kg
−
1
−1
kg and 167 kg kg respectively and represent the process
GAL. All processes can then be compared against these
figures. Other concepts such as average complexity, a trans-
a
Green Chemistry Centre of Excellence, Department of Chemistry, University of York,
formation GAL (tGAL) and relative process greenness (RPG) are
Heslington, York, YO10 5DD, UK. E-mail: rob.mcelroy@york.ac.uk
b
19–21
Organic Synthesis, Department of Chemistry, University of Antwerp,
also discussed.
With the availability of so many different
Groenenborgerlaan 171, 2020 Antwerp, Belgium. E-mail: bert.maes@uantwerpen.be
types of metrics, clarity was required to standardise how they
should be used as each metric has its advantages and disad-
vantages. A ‘Metrics Toolkit’ was developed at the University of
†
Electronic supplementary information (ESI) available: Appendix 1: summary of
experimental procedures and green metrics calculations; appendix 2: spread
sheet metrics calculations for amide synthesis; appendix 3: spread sheet metrics
calculations for Mitsunobu synthesis. See DOI: 10.1039/c9gc01537j
2
2
York, through the CHEM21 consortium. This toolkit takes a
holistic overview of several aspects of a process and allows a
‡
These authors contributed equally.
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 123

View Article Online
Paper
Green Chemistry
2
2
more rounded comparison of its credentials. On the other The PMI value can also be viewed in terms of its four inputs
hand, the toolkit does not tell the user if something is green (eqn (2)) or expressed as the amount of reagents, reactants and
but rather gives the information to determine if the process is catalyst (PMIRRC) and solvent (PMISolv) relative to the amount
better than the previous one and allows a direct comparison to of isolated product (eqn (3) and (4)).
be made between several processes. Importantly, it provides a
way to clearly identify which aspects can further be improved
to iteratively improve greenness.
total mass in a process or process step
PMI ¼
ð1Þ
mass of product
A direct comparison of physical reaction parameters yields
unambiguous data from which conclusions can be clearly
drawn, e.g., reaction temperature, solvent choice, the type of
work-up and purification employed (qualitative comparison
using flags). Similarly, mass-based metrics can also give unam-
biguous data e.g., percent yield, conversion, selectivity and
atom economy (quantitative comparison using figures). This is
massreactants þ massreagents þ masscatalyst þ masssolvent
PMI ¼
mass of isolated product
ð2Þ
massreactants þ massreagents þ masscatalyst
PMI ¼
mass of isolated product
ð3Þ
masssolvent
mass of isolated product
þ
2
3
illustrated in our recent paper, which compares a classical
and modern multi-component reaction (MCR) for the syn-
PMI ¼ PMIRRC þ PMISolv
ð4Þ
24–26
thesis of N-cyclohexyl-S-methyl-N′-phenylisothiourea.
example is particularly interesting as MCRs are considered to
be an inherently green methodology in the synthetic toolbox. pharmaceutical industry to measure the sustainability of a
The
PMI is the key metric adopted by the process chemists in
2
2
During this analysis the Chem21 Metrics Toolkit
was manufacturing process and was chosen for its ease in generat-
employed, and the results of a direct comparison between the ing, measuring and communicating data and ability to be
1
1
two processes are displayed in Table 1. All quantitative and used quickly to determine an estimation of greenness.
qualitative inputs should be considered as well as up-stream Another factor for selecting PMI as the metric of choice is
considerations such as how the reactants, reagents, and cata- due to its focus on process input rather than output, therefore
lysts have been prepared. Still, it is clear from these results driving innovation and efficiency from the outset. Moreover,
that some general assumptions can be made regarding yield, targets can be set and then used as a quantitative measure of
AE, RME (chosen to show the inherent efficiency with regard process performance/suitability. An example of this was pub-
2
7
to converting reactants into the desired product) and PMI (the lished by workers at Eli Lilly and Co. (Lilly). They reported
metric that accounts for all inputs in a step). Some of the that an original PMI target of <100 g g− was set for all APIs
results should have been expected whereas others such as in a commercial process, but soon discovered that trying to
RME may be considered to be surprising with respect to the meet a target <100 g g− was unrealistic and that it is more
general assumption about the inherent efficiency of these reac- suitable to adjust expectations according to the product
1
1
2
8
tions. From this work the metric which stands out the most to under study and its market demand. As a solution to this,
compare overall performance is the difference in PMI between workers at Lilly developed a strategy based on molecular com-
the two routes.
plexity and predicted market demand to set realistic PMI
2
7
PMI is the most complete mass-based metric as it com- targets.
prises the mass of all the material used in a synthetic route
There can be issues with setting limits based on a single
relative to the amount of isolated product (eqn (1)). Materials metric and also comparing mass-based metric results against
1
7
considered in its calculation include reagents, reactants, cata- different processes as briefly covered by Watson. A direct
lysts, solvents (reaction & purification) and work-up chemicals. comparison of the PMI between one reaction and another
23
Table 1 Quantitative metrics of classical and MCR approaches for the synthesis of N-cyclohexyl-S-methyl-N’-phenylisothiourea
2
4,25
Classical approach
2
6
MCR approach
PMI (g g−
1
)
PMIRRC (g g−1
a
)
PMISolv (g g−1
b
)
PMIWU (g g−1
c
)
Route
Yield (%)
AE (%)
RME (%)
Classical
MRC
84
88
66
76
53
39
817.1
324.5
26.9
13.8
742.3
312.5
790.2
310.7
a
b
c
Process mass intensity: reactants, reagents, catalyst. Process mass intensity: solvents. Process mass intensity: work-up.
124 | Green Chem., 2020, 22, 123–135
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
yielding the same product from the same reactants, but apply-
ing different reagents can appear to be a sound comparison at
Results and discussion
face value, but can the result reliably be compared without all Following on from these initial observations (Table 2), a
the (different) parameters of each reaction being taken into further in depth analysis on mass-based metrics was con-
consideration? The green credentials obtained via the Chem21 ducted. Amide bond formation is the largest category of reac-
Metrics Toolkit, of a reaction between piperazine (1) and tetra- tions carried out in the pharmaceutical industry, accounting
2
9–33
hydrofuran-2-carboxylic acid (2) via six different amide coup- for 16% of all transformations.
All practicing chemists will
ling reagents perfectly illustrates that issues can arise when inevitably encounter this reaction in their own research at one
metric targets are introduced and data directly compared stage in their career and therefore this is ideal as a model case.
(Table 2; for experimental details see ESI section 1†). The Several amides, taken from a comprehensive review on amide
highest PMI figure is surprisingly obtained from an enzymatic coupling reagents by Weisenburger et al., using a variety of
process, which may instinctively be the more favourable different coupling reagents were selected for our mass-based
3
1
process based on the reaction conditions (i.e. almost all green metrics analysis (Scheme 1). The top five coupling reagents
flags in Table 2). Moreover, the lowest PMI is obtained with in order of popularity (number of publications indicated
boric acid (B(OH) ) as coupling reagent although it also deli- between brackets underneath) in June 2015 were: 1-ethyl-3-(3′-
3
vers red flags based on the qualitative metrics. Why do these dimethylaminopropyl)carbodiimide hydrochloride (EDC; 54),
PMI values contradict our instincts as to which reaction is thionyl chloride (SOCl ; 42), carbonyldiimidazole (CDI; 38),
2
2
greener? Is it acceptable to just compare PMI figures without oxalyl chloride ((COCl) ; 22) and N,N′-dicyclohexylcarbo-
any caution when the same reactants are used under different diimide (DCC; 14) and therefore examples involving the appli-
reaction conditions? Moreover, can this also be done for two cation of these were chosen. The reagents pivaloyl chloride
different sets of reactants yielding different products? In this (PivCl; 13), isobutyl chloroformate (IBCF; 14) and propylpho-
3
manuscript, we seek answers to those questions. Of note, we sphonic anhydride (T P; 15) were also considered as coupling
will only be assessing quantitative mass based metrics, quali- reagents for our metrics analysis as a reasonable number of
tative metrics such as those covered by the Chem21 tool will processes employed them for reasons beyond the scope of this
3
1
not be considered. Fully assessing the greenness of any reaction paper (Scheme 1). As such, careful consideration of amide
requires a holistic approach. bond forming reactions should involve a broad range of reac-
Table 2 Comparison of the green credentials for amide bond formation from piperazine (1) and tetrahydrofuran-2-carboxylic acid (2) with six
22
coupling reagents using the CHEM21 Metrics Toolkit
Coupling agent
a
b
c
d
e
f
Silica
Enzyme
SOCl
2
B(OH)
3
PPh
3
& NBS
HMDS
Yield (%)
15
91
14
44
36
14
91
12
168
160
70
91
64
74
54
74
91
68
34
32
82
91
75
129
87
93
91
85
137
76
Atom economy (%)
RME (%)
−
1
PMI (g g
)
−
1
PMISolv (g g
)
Solvent choice
Catalyst?
Recoverable catalyst?
Critical element
Energy
Work-up
Health and safety
Chemical of concern?
—
—
—
—
—
—
—
—
2
2
Flag system: Green flag denotes preferred, amber is acceptable-some issues and red is undesirable. Reaction conditions: 1 (5 mmol), 2
a
b
(
5 mmol). K60 silica (0.1 g, activated at 700 °C), neat, 110 °C, 14 h, then acetone (20 mL). Immobilised novoenzyme 435 (0.1 g), heptane
c
d
(
20 mL), rt, 72 h. Thionyl chloride (15 mmol), toluene (20 mL), 110 °C, 1 h. Boric acid (0.5 mmol), toluene (20 mL), 110 °C, 11 h.
Triphenylphosphine (5 mmol), N-bromosuccinimide (5 mmol), pyridine (5 mmol), CH
e
f
2
Cl
2
(20 mL), 5 °C then rt, 1 h. Hexamethyldisilazane
(5 mmol), neat, 110 °C, 8 h.
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 125

View Article Online
Paper
Green Chemistry
31
Scheme 1 Overview of the selected amide bond forming reactions to evaluate the green metrics parameters.
tants and reagents to clearly highlight any issues arising from than actually required – especially in an early phase of develop-
a given metric. It will allow the results and trends generated to ment – and halogenated solvents such as dichloromethane,
be directly applicable to current discovery programs.
unfortunately, are still commonly used in extractions during
The experimental data of the selected reactions were input method development, although alternatives are in most cases
2
2
into the Chem21 Metrics Toolkit (see ESI†). The work-up of certainly possible. Considering PMI, including PMIWU, would
the assessed procedures was not taken into consideration as therefore give a distorted view not matching with reality with
the level of information given varied from publication to publi- respect to the chemical reaction itself. The parameters ana-
cation and its inclusion can easily present misleading metric lysed were percent yield, atom economy (AE), reaction mass
data when combined with the reagents and reactants in a efficiency (RME), process mass intensity (PMI, eqn (1) and (2)),
process. Moreover, typically larger volumes of solvent are taken process mass intensity for reagents, reactants & catalyst
126 | Green Chem., 2020, 22, 123–135
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
(PMIRRC, eqn (3) and (4)), and process mass intensity for reac- subsequently performed for both reactions to determine the
tion solvent (PMISolv, eqn (3) and (4)). PMIRRC was calculated impact of altering the reaction parameters on the green
by subtracting PMISolv from the overall process (PMI − metrics values. In simulation A both reactions were scaled to a
PMISolv). The concentration of the reaction was also calculated concentration of 0.4 M while keeping all other reaction para-
in relation to the limiting reactant. PMISolv and PMIRRC were meters (reactants, reagents, solvent, product, yield) unaltered
measured individually to highlight the impact each has on the compared to the literature reference, resulting in the PMI alter-
overall result. The dominance of solvents on the PMI is already ing considerably; in reaction 1 the PMI decreased from 17.3 to
1
1
−1
−1
known as shown by Manley et al. They described how sol- 5.7 g g and in reaction 2 the PMI increased from 3.9 g g to
vents normally account for 70% of the PMI. Using PMISolv and 14.6 g g− . When comparing both reactions by PMIRRC, the
PMIRRC allows the impact of solvents and reactants/reagents to solvent effect is ignored and the values of 1.8 g g & 2.2 g g
1
−
1
−1
be considered separately, therefore giving a clearer view of any remain constant (Table 3, simulation A). When the yield is
possible issues with and the variabilities in the data of the altered, the values for PMIRRC shift as seen in simulations B
reaction under study.
and D. As expected, decreasing the yield from 92% to 90% had
Table 3 contains the metrics data for two amide formation no effect on reaction 1. Increasing the yield from 81% to 90%
reactions (reactions 1 and 2), involving different reactants and in reaction 2 had only a minor change on the PMIRRC, from 2.2
) as coupling to 2.0 g g− . However, the effect was most noticeable when
1
products, but both with thionyl chloride (SOCl
2
reagent. The AE, RME and yield are very similar for both reac- there was a large change in yield to 50% (simulation D).
tions based on the literature data. However, a difference of Moreover, in this last simulation all green metrics values
3.4 g g− can be observed in the PMI of both reactions. A besides AE altered significantly compared to the reported lit-
1
1
−
1
similar difference (15.5 versus 1.7 g g ) is also noticeable in erature data for both reactions. This highlights the unreliabil-
the PMISolv, which is due to a large concentration difference, ity of comparing metrics from two reactions with significantly
0
.1 and 2.9 M respectively. The dominance of the solvent in differing parameters.
1
1
the PMI calculation, as reported by Manley et al. is again
Following the analysis and simulations with SOCl , atten-
2
nicely illustrated in this table. Four simulations (A–D) were tion was then turned to the other coupling reagents [oxalyl
a
Table 3 Literature data and simulation A–D with SOCl
2
as coupling reagent for reactions 1 and 2
Reaction 1
Reaction 2
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
AE (%)
RME (%)
Literature data reported
Reaction 1: [Acid] = 0.1 M
Reaction 2: [Acid] = 2.9 M
97
93
87
75
17.3
3.9
1.8
2.2
15.5
1.7
92
81
Simulation A: [Acid] = 0.4 M, literature yield
Reaction 1
Reaction 2
97
93
87
75
5.7
14.6
1.8
2.2
3.9
12.4
92
81
Simulation B: [Acid] = literature data, yield = 90%
Reaction 1
Reaction 2
97
93
85
83
17.7
3.5
1.8
2.0
15.8
1.5
90
90
Simulation C: [Acid] = 0.4 M, yield = 90%
Reaction 1
Reaction 2
97
93
85
83
5.8
13.2
1.8
2.0
4.0
11.2
90
90
Simulation D: [Acid] = 0.4 M, yield = 50%
Reaction 1
Reaction 2
97
93
47
46
10.5
23.7
3.3
3.5
7.2
20.2
50
50
a
Reactions refer to Scheme 1.
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 127

View Article Online
Paper
Green Chemistry
chloride (Table S1†), CDI (Table S2†), DCC (Table S3†), EDC molecular weight difference between reactants gives the lowest
(
(
Table S4†), PivCl (Table S5†), IBCF (Table S6†), and T3P PMI/PMIRRC values and the lowest difference the reverse, though
Table S7†)], based on the selected literature examples shown the yield of the target amide is the same. This conclusion
3
1
in the review (Scheme 1). All target compounds are amides remains when exactly the same coupling agent (CDI or IBCF) and
with a variety of molecular weights between 220.27 and stoichiometry (excess coupling reagent and amine reactant) were
97.88 g mol− . The same trends for yield and concentration applied for reactions 5 and 13 (simulations F–I).
1
8
on PMI and PMIRRC as observed in Table 3 were also con-
Quantitative green metrics assessment is also interesting for
firmed throughout the analysis of this data. The lower the con- methodology development – though not performed yet – when
centration used in the reaction, the higher the PMI. This is researchers want to benchmark their synthetic method with
logical since a concentrated process will contain less solvent. the existing literature. The effect of alteration of the coupling
This is especially significant when comparing processes, e.g. agent for the synthesis of the same amide reaction product
pilot plant/production scale versus a discovery route. The con- was therefore evaluated. This was again examined for reactions
ditions used during bench work applying standard laboratory 5 and 13 with a set of different coupling reagents covering a
techniques and dilutions are notably different to those larger MW range (between 118.97 g mol− and 318.18 g mol
1
−1
;
3
4
employed at pilot or production scale. Reaction concen- Table 5). Yield, concentration and excess reagent were fixed in
tration is a factor which will be considered and optimized, these simulations. The PMIRRC ranges from 1.6–2.2 g g (reac-
−1
−1
most likely by a process chemist, whereas at the bench in a tion 5) and from 1.5–1.8 g g (reaction 13) and the overall PMI
high throughput R&D laboratory, concentration is rarely con- from 8.0–8.6 g g− (reaction 5) and from 4.3–4.6 g g
1
−1
sidered and typically varies a lot. A direct comparison of two (reaction 13). Therefore, the effect of the coupling reagent on the
reactions here would therefore most likely be inaccurate, PMI/PMIRRC is small. On the other hand, when there is a large
imposing the use of PMIRRC. Even when using PMIRRC simu- span in MW of the reactants, when reactions 5 and 13 are com-
lations with similar yield will be required to compare the green pared, the PMI will be lower than when their MWs are more
potential of amide bond-forming reactions differing in product equal. When there is a large span in MW of the reactants, the
and coupling reagent.
PMI will be higher than when their MWs are more equal as
While in Table 3 two reactions with SOCl
2
as coupling observed before (vide supra). It is interesting to note the range of
reagent were randomly selected, we subsequently compared PMIRRC is however small. The effect on PMI is more dependent
two amide bond-forming reactions of Scheme 1 featuring both on the reactants than the reagents and this should be taken into
−
1
the smallest (reaction 5, 17.92 g mol ) and largest (reaction consideration when developing a synthetic procedure.
3, 461.36 g mol− ) molecular weight difference between the
1
Looking at the best yielding reactions from the eight
1
two reactants (Table 4). Obviously, this is also reflected in the different coupling reagents for amide synthesis covered in this
molecular weight of the corresponding products (388.46 g paper a number of trends become clear (Table 6). The overall
−
1
−1
mol versus 897.88 g mol ). Though the coupling reagent is yields are high which is expected from an optimised process –
not the same (CDI and IBCF), the molecular weight difference the best being quantitative and the worst 84% – allowing
−1
is small (23.57 g mol ). When comparing the literature reliable comparison. However, in order to eliminate the con-
metrics for both processes there is again very little difference centration effect, the concentration of all reactions was again
between them at first glance; the yield 89% vs. 84%, PMI 8.8 scaled to 0.4 M. The AEs are also high figures ranging from
−
1
−1
vs. 8.3 g g , PMIRRC 2.2 vs. 2.0 g g and PMISolv 6.6 vs. 6.3 g 97–83%. A slight spread appears in RME with a range of
−
1
−1
g
0
. However, when the concentrations were standardised at 96–67%. PMI has a range from 15.0–4.9 g g which are rela-
.4 M the difference in PMI became apparent; the PMI of 8.8 tively good results for a synthetic process. The highest and
−
1
−1
−1
vs. 8.3 g g went to 8.7 vs. 4.9 g g for both reactions (simu- lowest PMIRRC are 3.3 & 1.8 g g , respectively. As mentioned
lation A). When this effect is combined with a yield of only previously PMIRRC is a small proportion of the overall metric
0% the values went to 15.5 vs. 8.3 g g− (simulation D). This but looking at Table 6 one would argue that the reaction with
1
5
clearly highlights the problem of comparing a PMI value the lowest PMIRRC should therefore have the greenest potential
without standardising the concentration. A concentration of irrespective of other factors which can influence selection.
0
.4 M was chosen as the default concentration as we felt it This would indicate thionyl chloride as being one of the best
reflected a reasonable concentration that a working bench choices for amide coupling, which is actually in line with the
chemist may arbitrarily use. It is worth noting that the chosen OPRD review and supports the fact that thionyl chloride is by
concentration is somewhat irrelevant as long as it is standardised. far one of the most popular choices for amide
2
9,30–32
In simulation C the concentration is 0.4 M whereas in simulation formation.
E the concentration has been scaled to 0.8 M, both with a yield of factors such as its low cost, ease of use and the gaseous by-pro-
0%. As expected, in both simulations the values for PMIRRC are ducts eliminating any extra separation step and again helping
This is besides metrics also due to other
9
identical making it a reliable metric for comparison. The overall to reduce the cost of the process. Based on Table 6, PMIRRC
PMI is highest for the lowest concentration (simulation C) as alone seems to be a viable alternative to the full PMI calcu-
would be expected. Standardising the concentration avoids the lation for reaction conditions selection (i.e., reagents). This
effect that the quantity of solvent has on the overall metrics of the would allow R&D chemists to truly see the advantages or disad-
process. It is interesting to note that the reaction with the highest vantages of a particular approach although concentration and
128 | Green Chem., 2020, 22, 123–135
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
Table 4 Yield and concentration simulations for reactions with smallest (reaction 5) and largest (reaction 13) molecular weight difference between
a
the reactants
ΔMW reactants 17.92 g mol−
1
MW CDI 162.15 g mol−
1
Reaction 5 (CDI)
ΔMW reactants 461.36 g mol−
1
MW IBCF 138.58 g mol−
1
Reaction 13 (IBCF)
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
AE (%)
RME (%)
Literature data reported
Reaction 5: [Acid] = 0.4 M
Reaction 13: [Acid] = 0.2 M
96
83
81
67
8.8
8.3
2.2
2.0
6.6
6.3
89
84
Simulation A: [Acid] = 0.4 M, literature yield
Reaction 5
Reaction 13
96
83
81
67
8.7
4.9
2.2
2.0
6.4
2.9
89
84
Simulation B: [Acid] = literature data, yield = 90%
Reaction 5
Reaction 13
96
83
82
72
8.6
7.8
2.2
1.9
6.4
5.9
90
90
Simulation C: [Acid] = 0.4 M, yield = 90%
Reaction 5
Reaction 13
96
83
82
72
8.6
4.6
2.2
1.9
6.4
2.7
90
90
Simulation D: [Acid] = 0.4 M, yield = 50%
Reaction 5
Reaction 13
96
83
46
40
15.5
8.3
4.0
3.4
11.5
4.9
50
50
Simulation E: [Acid] = 0.8 M, yield = 90%
Reaction 5
Reaction 13
96
83
82
72
5.4
3.2
2.2
1.9
3.2
1.4
90
90
Simulation F: CDI (1.1 eq.) as coupling reagent for both reactions, reactant stoichiometry as literature, [acid] = 0.4 M, Yield = 90%
Reaction 5
96
83
82
72
8.6
4.5
2.2
1.8
6.4
2.7
90
90
Reaction 13
Simulation G: CDI (1.1 eq.) as coupling reagent for both reactions, reactant stoichiometry as literature, [acid] = 0.4 M, Yield = 50%
Reaction 5
96
83
46
40
15.4
8.2
3.9
3.3
11.5
4.9
50
50
Reaction 13
Simulation H: IBCF (1.1 eq.) as coupling reagent for both reactions, reactant stoichiometry as literature, [acid] = 0.4 M, Yield = 90%
Reaction 5
96
83
82
72
8.4
4.5
2.0
1.7
6.4
2.8
90
90
Reaction 13
Simulation I: IBCF (1.1 eq.) as coupling reagent for both reactions, reactant stoichiometry as literature, [acid] = 0.4 M, Yield = 50%
Reaction 5
96
83
46
40
15.1
8.0
3.5
3.1
11.5
4.9
50
50
Reaction 13
a
Reactions refer to Scheme 1.
solvent usage would need to be considered separately after- an amine. If for example, conditions A & B where analysed and
wards. The PMIRRC value is generally small compared to PMIRRC for B generates a lower value, it would indicate an
PMISolv and PMIWU, therefore using this figure to compare two or inherent organic reaction-based improvement to the process of A.
more coupling agents would highlight the chemistry, i.e. reagent
s) selected, with the greatest green potential to perform the the viability of green chemistry metrics for the Mitsunobu
specific reaction, e.g. amide synthesis from a carboxylic acid and esterification reaction, chosen because of its widespread use in
In addition to amide bond formation, we also investigated
(
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 129

View Article Online
Paper
Green Chemistry
Table 5 Simulations for 8 different coupling reagents for reactions with smallest (reaction 5) and largest (reaction 13) molecular weight difference
a
between the reactants. Reactant stoichiometry as described in literature for all simulations
ΔMW reactants 17.92 g mol−
1
Reaction 5
ΔMW reactants 461.36 g mol−
1
Reaction 13
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
AE (%)
RME (%)
1
Simulation A: SOCl
Reaction 5
Reaction 13
2
(1.1 eq., MW = 118.97 g mol− ), [Acid] = 0.4 M in THF
96
83
82
72
8.0
4.3
1.6
1.5
6.4
2.8
90
90
1
Simulation B: (COCl)
Reaction 5
Reaction 13
2
(1.1 eq., MW = 126.93 g mol− ), [Acid] = 0.4 M in THF
96
83
82
72
8.0
4.3
1.6
1.6
6.4
2.8
90
90
Simulation C: CDI (1.1 eq., MW = 162.15 g mol− ), [Acid] = 0.4 M in THF
1
Reaction 5
96
83
82
72
8.1
4.4
1.7
1.6
6.4
2.8
90
90
Reaction 13
Simulation D: DCC (1.1 eq., MW = 206.33 g mol− ), [Acid] = 0.4 M in THF
1
Reaction 5
96
83
82
72
8.3
4.4
1.9
1.7
6.4
2.8
90
90
Reaction 13
Simulation E: EDC (1.1 eq., MW = 191.70 g mol− ), [Acid] = 0.4 M in THF
1
Reaction 5
96
83
82
72
8.2
4.4
1.8
1.6
6.4
2.8
90
90
Reaction 13
Simulation F: PivCl (1.1 eq., MW = 120.58 g mol− ), [Acid] = 0.4 M THF
1
Reaction 5
96
83
82
72
8.0
4.3
1.6
1.5
6.4
2.8
90
90
Reaction 13
Simulation G: IBCF (1.1 eq., MW = 138.58 g mol− ), [Acid] = 0.4 M in THF
1
Reaction 5
96
83
82
72
8.0
4.3
1.6
1.6
6.4
2.8
90
90
Reaction 13
1
Simulation H: T
Reaction 5
Reaction 13
3
P (1.1 eq., MW = 318.18 g mol− ), [Acid] = 0.4 M in THF
96
83
82
72
8.6
4.6
2.2
1.8
6.4
2.8
90
90
a
Reactions refer to Scheme 1.
a
Table 6 Overview of the amide coupling reagents. All values (yield, stoichiometry) as reported in literature besides concentration (scaled to 0.4 M)
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
AE (%)
RME (%)
Reaction 1 (SOCl
2
)
97
95
96
91
87
94
83
96
87
79
96
77
79
94
67
81
5.7
12.2
8.7
6.7
8.5
15.0
4.9
8.1
1.8
3.2
2.2
2.1
2.6
2.3
2.0
3.3
3.9
9.0
6.4
4.5
5.9
12.6
2.9
4.7
92
91
89
86
92
100
84
88
Reaction 3 ((COCl)
Reaction 5 (CDI)
Reaction 7 (DCC)
Reaction 9 (EDC)
Reaction 11 (PivCl)
Reaction 13 (IBCF)
2
)
Reaction 14 (T
3
P)
a
Reactions refer to Scheme 1.
130 | Green Chem., 2020, 22, 123–135
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
pharmaceutical industry. Synthetic interest is based on its use was not found to have a major impact in amide synthesis. It is
for the selective inversion of a secondary alcohol’s chirality important to note that in the amide synthesis additives
3
5
while converting the functionality to an ester. It employs two (mostly bases, nucleophilic activators) are often used which of
stoichiometric coupling reagents (triphenylphosphine and course also belong to the reagent category for green metrics
diethyl azodicarboxylate) opposed to one in the previous case analysis. However, simulations with additives that are actually
study. This makes it attractive for green metrics analysis as it practically used, spanning a large difference in molecular
allows us to evaluate the effect of multiple reagents on the weight, are not self-evident to identify here. Like before, the
parameters and to assess if our previous trends and obser- reaction conditions were inserted into the Chem21 Metrics
2
2
vations are still valid, especially considering reagent variation Toolkit and a variety of simulations analysed (Table 7).
Table 7 Green metrics simulations applied on the Mitsunobu reaction of benzyl alcohol with various carboxylic acids (A–E) featuring a different
molecular weight
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
Carboxylic acid
AE (%)
RME (%)
Simulation 1: [Acid] = 0.4 M, yield = 90%
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
15.7
13.7
11.4
8.8
4.1
3.6
3.2
2.7
2.3
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
6.9
Simulation 2: [Acid] = 0.4 M, yield = 80%
A
B
C
D
E
92
93
94
96
97
67
69
71
73
74
17.7
15.4
12.8
9.9
4.6
4.1
3.6
3.0
2.6
13.1
11.3
9.2
6.9
5.2
80
80
80
80
80
7.8
Simulation 3: [Acid] = 0.4 M, yield = 70%
A
B
C
D
E
92
93
94
96
97
59
60
62
64
65
20.2
17.6
14.6
11.4
8.9
5.2
4.7
4.1
3.4
2.9
15.0
12.9
10.5
7.9
70
70
70
70
70
6.0
Simulation 4: [Acid] = 0.4 M, yield = 50%
A
B
C
D
E
92
93
94
96
97
42
43
44
46
47
28.3
24.6
20.4
15.9
12.5
7.3
6.6
5.7
4.8
4.1
20.9
18.0
14.7
11.1
8.4
50
50
50
50
50
Simulation 5: Scale reaction × 5: [Acid] = 0.4 M, yield = 90%
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
15.7
13.7
11.4
8.8
4.1
3.6
3.2
2.7
2.3
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
6.9
Simulation 6: [Acid] = 0.8 M, yield = 90%
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
9.9
8.6
7.3
5.7
4.6
4.1
3.6
3.2
2.7
2.3
5.8
5.0
4.1
3.1
2.3
90
90
90
90
90
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 131

View Article Online
Paper
Green Chemistry
Table 8 Green metrics simulations applied on the Mitsunobu reaction of benzyl alcohol with two carboxylic acids altering the AZO-coupling
reagent for reactions with a small (A) and large (E) molecular weight difference between the reactants
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
Carboxylic acid
AE (%)
RME (%)
Simulation 1: [Acid] = 0.4 M, yield = 90%, DEAD as coupling reagent
A
E
92
97
76
84
15.7
6.9
4.1
2.3
11.6
4.7
90
90
Simulation 2: [Acid] = 0.4 M, yield = 90%, DCPEAC as coupling reagent
A
E
92
97
76
84
16.2
7.1
4.5
2.5
11.6
4.7
90
90
Simulation 3: [Acid] = 0.4 M, yield = 90%, DBPEAC as coupling reagent
A
E
92
97
76
84
16.7
7.4
5.1
2.7
11.6
4.7
90
90
Benzyl alcohol was coupled with several carboxylic acids (A–E) boxylic acid was examined (Table 8). Triphenylphosphine
with molecular weights ranging from 122.12 to 440.50 g mol−
1
.
(262.29 g mol ) was combined with diethyl azodicarboxylate
−1
The simulations considered variations in yield, concentration (DEAD, 174.16 g mol− ), ethyl 2-(2,3-dichlorophenyl)diazene-1-
1
and excess of reagents as before. The full set of simulations for carboxylate (DCPEAC, 247.08 g mol− ) or ethyl 2-(2,3-dibromo-
1
−
1
the Mitsunobu reaction can be found in the ESI section 3.† phenyl)diazene-1-carboxylate (DBPEAC, 335.98 g mol ). Yield,
Comparison of the same reaction of benzyl alcohol with concentration and excess reagent were again fixed in these
−
1
benzoic acid at the same concentration, but with varying yield simulations as before. The PMIRRC ranges from 4.1–5.1 g g
−
1
is shown in Table 7 (simulations 1–4 carboxylic acid A). The (A) and from 2.3–2.7 g g (E) and the overall PMI from
PMI varies from 15.7 g g to 28.3 g g for a yield difference 15.7–16.7 g g (A) and from 6.9–7.4 g g−1 (E). Similar to amide
−
1
−1
−1
of 40%. Altering the concentration from 0.4 M to 0.8 M for synthesis the effect on PMI/PMIRRC is more dependent on the
9
1
0% yield varies the PMI with a value of 5.8 g g−1 (simulations reactants than the reagents, irrespective of the number of
and 6, carboxylic acid A). Yield and concentration therefore reagents used. When comparing two processes it intrinsically
have a similar effect on the PMI/PMIRRC as observed in the may be logical to assume the process employing only one
amide coupling. Again, if one is not careful in the selection of reagent would perform more favourable than the process
the correct example to determine the greenness of a reaction, using two. This may not be the case when looking at PMI as
one could draw the conclusion that the coupling between only the mass is considered and the sum of two moderate
benzyl alcohol and benzoic acid (A) with a PMI of 15.7 g g− is weight reagents may actually be lower than the mass of one
1
less favourable than the coupling between benzyl alcohol and bulky metal complex!
an amino-protected ornithine (E) with a PMI of 6.9 g g−
(
1
Finally, we also evaluated the recent advances towards a
3
6,37
Table 7, simulation 1) simulated at the same yield, concen- catalytic Mitsunobu reaction
and compared these with
tration and reaction stoichiometry. However, this does not classical alcohol activation at the expense of stoichiometric
appreciate the green potential of the reaction itself as this phosphine an azodicarboxylate coupling reagents (Table 9).
difference in PMI is just dependent on the MW of the reactants Recent studies recycle the required stoichiometric reagents
3
8,39
as observed in amide coupling. This again highlights the in situ using terminal oxidants [di(acetoxy)iodobenzene,
4
0,41
potential pitfall when comparing the PMI of different sets of
reactants for discovery purposes and how, based on the reac- lane ). However, a stoichiometric reagent will always be
tants being studied, a favourable PMI can be obtained. required to allow the phosphine or azodicarboxylate com-
O
2
with iron phthalocyanine
] and reductants (phenylsi-
4
2
The effect of the molecular weight of the reagents for the ponents to be used in catalytic amounts. Therefore, any
coupling of benzyl alcohol with a low (benzoic acid, A) and improvements in green metrics are explicitly linked to the
high (amino-protected ornithine, E) molecular weight car- nature of the terminal oxidants and reductants employed. For
132 | Green Chem., 2020, 22, 123–135
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
Table 9 Green metrics simulations applied on the (catalytic) Mitsunobu reaction of benzyl alcohol with various carboxylic acids (A–E) featuring a
different molecular weight
PMI (g g−
1
)
PMIRRC (g g−1
)
PMISolv (g g−1
)
Yield (%)
Carboxylic acid
AE (%)
RME (%)
Simulation 1: Stoichiometric coupling reagents [PPh
3
(1.2 eq.), DCPEAC (1.2 eq.)]
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
16.2
14.1
11.7
9.1
4.5
4.0
3.5
2.9
2.5
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
7.1
3 2
Simulation 2: Catalytic in azodicarboxylate coupling reagent [PPh (1.2 eq.), DCPEAC (0.1 eq.), Fe(Pc) (0.1 eq.), O ]
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
15.0
13.1
10.9
8.5
3.4
3.1
2.7
2.3
2.0
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
6.7
Simulation 3: Catalytic in azodicarboxylate coupling reagent [PPh
3
2
(1.2 eq.), DCPEAC (0.1 eq.), PhI(OAc) (2.0 eq.)]
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
18.1
15.7
13.1
10.1
7.9
6.5
5.7
4.9
4.0
3.3
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
3
Simulation 4: Catalytic in phosphine coupling reagent [1-phenyl phospholane (0.1 eq.), DCPEAC (1.2 eq.), PhSiH (1.1 eq.)]
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
15.2
13.3
11.0
8.6
3.6
3.3
2.9
2.4
2.1
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
6.8
Simulation 5: ‘Fully catalytic’ [PPh
3
(0.1 eq.), DCPEAC (0.1 eq.), Fe(Pc) (0.1 eq.), O
2
, PhSiH
3
(1.1 eq.)]
A
B
C
D
E
92
93
94
96
97
76
78
80
82
84
14.1
12.3
10.2
8.0
2.5
2.3
2.1
1.8
11.6
10.0
8.2
6.2
4.7
90
90
90
90
90
6.3
1.7
example, the highest PMIRRC is obtained in simulation 3 of
Table 9, which can be explained by two equivalents of
Conclusions
di(acetoxy)iodobenzene that are required for re-oxidation; In this study into mass-based metrics, we have highlighted the
thereby generating a large amount of waste. In contrast with potential pitfalls of introducing metric targets to select an
O
2
and catalytic iron phthalocyanine a lower PMIRRC is optimal synthetic procedure at discovery level. By evaluating
obtained (simulation 2). In order to improve the PMI, Aldrich the green potential of amide bond forming and Mitsunobu
and Taniguchi reported a protocol that is catalytic in both reactions, it has been illustrated that process mass intensity
phosphine (1-phenylphospholane-1-oxide) and azocarboxylate (PMI, or PMIRRC) can be used as a very effective metric for
employing phenylsilane reductant to recycle the phosphine reagents selection. However, one needs to take care when com-
catalyst and iron-phthalocyanine/O
2
oxidant to regenerate the paring different reactions with respect to green metrics.
4
2,43
azocarboxylate.
Pleasingly the PMIRCC is indeed lower than Different simulations have highlighted how PMI results could
for the stoichiometric coupling (simulations 1 and 5, Table 9).
be misleading when comparing values at bench scale without
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 133

View Article Online
Paper
Green Chemistry
a due consideration of all the relevant parameters. In order Seventh Framework Programme (FP7/2007-2013) and EFPIA
to make a more accurate comparison between different pro- companies in kind contribution.
cedures, the yield, concentration (unless PMIRCC is con-
sidered) and the difference in molecular weights of the reac-
tants should be as close as practically possible. Of course
not all protocols will practically allow certain transform-
References
ations to be performed at the same concentration.
Interestingly, our results show that the effect of the mole-
cular weight of the applied reagent is rather limited.
Practically, when researchers want to benchmark their syn-
thetic method with existing literature for the same reaction,
it is extremely important to be very careful in the selection
of the examples used for comparison as difference in PMI
should not reflect the difference in molecular weight of the
reactants/products used for comparative analysis, but rather
identify which methodologies have the greatest green poten-
tial. As such, those methodologies should be focused on for
optimisation in a discovery context and not abandoned due
to initial poor data obtained from the bench or from the lit-
erature, avoiding a fair green metrics comparison. Easy to
perform simulations are crucial in this respect as high-
1 R. A. Sheldon, Chem. Ind., 1997, 12–15.
2 R. A. Sheldon, Green Chem., 2007, 9, 1273–1283.
3 Green Techniques for Organic Synthesis and Medicinal
Chemistry, ed. W. Zhang and B. W. Cue, Wiley, 2nd edn,
2012.
4 D. J. C. Constable, A. D. Curzons and V. L. Cunningham,
Green Chem., 2002, 4, 521–527.
5 J. A. Glaser, Clean Technol. Environ. Policy, 2009, 11, 371.
6 P. J. Dunn, Chem. Soc. Rev., 2012, 41, 1452–1461.
7 D. J. C. Constable, P. J. Dunn, J. D. Hayler, G. R. Humphrey,
J. J. L. Leazer, R. J. Linderman, K. Lorenz, J. Manley,
B. A. Pearlman, A. Wells, A. Zaks and T. Y. Zhang, Green
Chem., 2007, 9, 411–420.
8 B. Trost, Science, 1991, 254, 1471–1477.
9 R. A. Sheldon, Chem. Tech., 1994, 24, 38–47.
lighted in this work. Since the synthetic community moves 10 A. D. Curzons, D. J. C. Constable, D. N. Mortimer and
towards greener approaches for the synthesis of industrially V. L. Cunningham, Green Chem., 2001, 3, 1–6.
relevant compounds, our study may serve as a good guide to 11 C. Jimenez-Gonzalez, C. S. Ponder, Q. B. Broxterman and
correctly apply PMI as an effective discovery level metric. J. B. Manley, Org. Process Res. Dev., 2011, 15, 912–917.
One should of course still bear in mind that additional 12 C. Jiménez-González, C. Ollech, W. Pyrz, D. Hughes,
qualitative factors, as stated in the original Chem21 toolkit
paper, also need to be taken into consideration such as
Q. B. Broxterman and N. Bhathela, Org. Process Res. Dev.,
2013, 17, 239–246.
2
2
elemental sustainability, renewability, and health and safety 13 C. Jiménez-González, D. J. Constable, A. D. Curzons and
of reagents/reactants/solvents. Though PMI is an important
metric, no metric should be assessed in isolation, but rather
V. L. Cunningham, Clean Technol. Environ. Policy, 2002, 4,
44–53.
in a holistic manner to assess the overall green credentials of 14 C. Jiménez-González, D. J. C. Constable and C. S. Ponder,
a synthetic methodology. Chem. Soc. Rev., 2012, 41, 1485–1498.
The findings of this paper have been applied in a review 15 G. J. Ruiz-Mercado, R. L. Smith and M. A. Gonzalez, Ind.
4
4
discussing different thiosulfonate syntheses. Here incorpor-
ation of the green credentials of the synthetic procedures 16 M. G. T. C. Ribeiro and A. A. S. C. Machado, Green Chem.
towards thiosulfonates allows to judge the merits of the state Lett. Rev., 2013, 6, 1–18.
of the art, beyond the typical yield of a product and availability 17 F. I. McGonagle, H. F. Sneddon, C. Jamieson and
Eng. Chem. Res., 2012, 51, 2309–2328.
of the reactants.
A. J. B. Watson, ACS Sustainable Chem. Eng., 2014, 2, 523–532.
8 L. Summerton and A. Constandinou, Beyond Mass-based
Metrics: Evaluating the Greenness of Your Reaction in Green
and Sustainable Medicinal Chemistry: Methods, Tools and
Strategies for the 21st Century Pharmaceutical Industry, The
Royal Society of Chemistry, 2016, pp. 41–53.
1
Conflicts of interest
The authors declare no conflicts of interest.
1
2
9 F. Roschangar, R. A. Sheldon and C. H. Senanayake, Green
Chem., 2015, 17, 752–768.
0 F. Roschangar, J. Colberg, P. J. Dunn, F. Gallou,
J. D. Hayler, S. G. Koenig, M. E. Kopach, D. K. Leahy,
I. Mergelsberg, J. L. Tucker, R. A. Sheldon and
C. H. Senanayake, Green Chem., 2017, 19, 281–285.
Acknowledgements
This research was financially supported by the University of
Antwerp (BOF), the Hercules and Francqui Foundation, the
Research Foundation Flanders (FWO) (Research Project, 21 F. Roschangar, Y. Zhou, D. J. C. Constable, J. Colberg,
Scientific Research Community, BioFact Excellence of Science
project Grant no. 30902231) and the Innovative Medicines
Initiative (http://www.imi.europa.eu) Joint Undertaking under
grant agreement no. 115360, resources of which are composed
of the financial contribution from the European Union’s
D. P. Dickson, P. J. Dunn, M. D. Eastgate, F. Gallou,
J. D. Hayler, S. G. Koenig, M. E. Kopach, D. K. Leahy,
I. Mergelsberg, U. Scholz, A. G. Smith, M. Henry, J. Mulder,
J. Brandenburg, J. R. Dehli, D. R. Fandrick, K. R. Fandrick,
F. Gnad-Badouin, G. Zerban, K. Groll, P. T. Anastas,
134 | Green Chem., 2020, 22, 123–135
This journal is © The Royal Society of Chemistry 2020

View Article Online
Green Chemistry
Paper
R. A. Sheldon and C. H. Senanayake, Green Chem., 2018, 34 http://learning.chem21.eu/process-design/-/introduction-to-
0, 2206–2211. process-engineering/what-is-scale-up.
2 C. R. McElroy, A. Constantinou, L. C. Jones, L. Summerton 35 S. Fletcher, Org. Chem. Front., 2015, 2, 739–752.
2
2
2
2
2
2
and J. H. Clark, Green Chem., 2015, 17, 3111–3121.
3 S. Abou-Shehada, P. Mampuys, B. U. W. Maes, J. H. Clark
and L. Summerton, Green Chem., 2017, 19, 249–258.
4 C. Levallet, J. Lerpiniere and S. Y. Ko, Tetrahedron, 1997,
36 R. H. Beddoe, H. F. Sneddon and R. M. Denton, Org.
Biomol. Chem., 2018, 16, 7774–7781.
37 D. Hirose, M. Gazvoda, J. Košmrlj and T. Taniguchi, Chem.
Sci., 2016, 7, 5148–5159.
5
3, 5291–5304.
38 T. Y. S. But and P. H. Toy, J. Am. Chem. Soc., 2006, 128,
9636–9637.
5 S. K. Hamilton, D. E. Wilkinson, G. S. Hamilton and
Y.-Q. Wu, Org. Lett., 2005, 7, 2429–2431.
39 T. Y. S. But, J. Lu and P. H. Toy, Synlett, 2010, 1115–1117.
6 P. Mampuys, Y. Zhu, T. Vlaar, E. Ruijter, R. V. A. Orru and 40 D. Hirose, T. Taniguchi and H. Ishibashi, Angew. Chem.,
B. U. W. Maes, Angew. Chem., Int. Ed., 2014, 53, 12849– Int. Ed., 2013, 52, 4613–4617.
2854. 41 O has been neglected in the PMI calculations given its low
1
2
2
7 D. P. Kjell, I. A. Watson, C. N. Wolfe and J. T. Spitler, Org.
Process Res. Dev., 2013, 17, 169–174.
molecular weight. It is the most abundant and sustainable
oxidant on earth yielding water as a by-product also justify-
ing its omission from the calculations. For a review describ-
2
2
8 W. J. W. Watson, Green Chem., 2012, 14, 251–259.
9 S. D. Roughley and A. M. Jordan, J. Med. Chem., 2011, 54,
ing important parameters when using O , see: H. Sterckx,
2
3
451–3479.
B. Morel and B. U. W. Maes, Angew. Chem., Int. Ed., 2019,
58, 7946–7970.
42 J. A. Buonomo and C. C. Aldrich, Angew. Chem., Int. Ed.,
2015, 54, 13041–13044.
43 D. Hirose, M. Gazvoda, J. Košmrlj and T. Taniguchi, Org.
Lett., 2016, 18, 4036–4039.
3
3
3
3
0 R. M. de Figueiredo, J.-S. Suppo and J.-M. Campagne,
Chem. Rev., 2016, 116, 12029–12122.
1 J. R. Dunetz, J. Magano and G. A. Weisenburger, Org.
Process Res. Dev., 2016, 20, 140–177.
2 K. Hollanders, B. U. W. Maes and S. Ballet, Synthesis, 2019,
5
1, 2261–2277.
44 P. Mampuys, C. R. McElroy, J. H. Clark, R. V. A. Orru and
B. U. W. Maes, Adv. Synth. Catal., 2009, DOI: 10.1002/
adsc.201900864.
3 M. T. Sabatini, L. T. Boulton, H. F. Sneddon and
T. D. Sheppard, Nat. Catal., 2019, 2, 10–17.
This journal is © The Royal Society of Chemistry 2020
Green Chem., 2020, 22, 123–135 | 135