DOI: 10.1002/chem.

201204558

A Continuous-Flow Process for the Synthesis of Artemisinin

Daniel Kopetzki,[a] FranÅois Lvesque,[a] and Peter H. Seeberger*[a, b]

Abstract: Isolation of the most effec-
tive antimalarial drug, artemisinin,
from the plant sweet wormwood, does
not yield sufficient quantities to pro-
vide the more than 300 million treat-
ments needed each year. The high
prices for the drug are a consequence
of the unreliable and often insufficient
supply of artemisinin. Large quantities
of ineffective fake drugs find a market
in Africa. Semisynthesis of artemisinin
from inactive biological precursors,
either dihydroartemisinic acid
(DHAA) or artemisinic acid, offers a
potentially attractive route to increase
artemisinin production. Conversion of
the plant waste product, DHAA, into
artemisinin requires use of photochem-
ically generated singlet oxygen at large
scale. We met this challenge by devel-
oping a one-pot photochemical contin-
uous-flow process for the semisynthesis
of artemisinin from DHAA that yields
Keywords: antimalarial agents · ar-
temisinin · flow chemistry · photo-
chemistry · singlet oxygen
65 % product. Careful optimization re-
sulted in a process characterized by
short residence times. A method to ex-
tract DHAA from the mother liquor
accumulated during commercial arte-
misinin extractions, a material that is
currently discarded as waste, is also re-
ported. The synthetic continuous-flow
process described here is an effective
means to supplement the limited avail-
ability of artemisinin and ensure in-
creased supplies of the drug for those
in need.

Introduction misinin is still exclusively obtained by extraction from the

Artemisinin combination therapies (ACTs) are currently the
most effective way to combat malaria.[1] By far the most ex-
pensive and thus limiting ingredient in ACTs is artemisi-
nin.[2] It was first discovered in the 1970 s as part of a con-
certed effort in China to identify new antimalarial agents by
combing through the ancient literature of traditional Chi-
nese medicine and screening hundreds of herbal extracts.[3]
An extract of sweet wormwood (Artemisia annua) contained
a promising antimalarial candidate. Finally, the structure of
the active ingredient that was called artemisinin was eluci-
dated, thus revealing a complex endoperoxide skeleton that
proved to be responsible for the activity against plasmodium
parasites.[4] Artemisinin as well as its derivatives that exhibit
improved uptake and activity can clear most of the parasites
in just a few hours.[5]
The total synthesis of artemisinin, owing to the molecular
complexity of the structure, is commercially not viable even
though several routes have been reported.[6] Therefore, arte-
[a] Dr. D. Kopetzki, Dr. F. Lvesque, Prof. Dr. P. H. Seeberger
Department for Biomolecular Systems
Max Planck Institute of Colloids and Interfaces
14424 Potsdam (Germany)
Fax: (+ 49) 331-567-9302
E-mail: peter.seeberger@mpikg.mpg.de
[b] Prof. Dr. P. H. Seeberger
Institute for Chemistry and Biochemistry
Freie Universitt Berlin
Arnimallee 22, 14195 Berlin (Germany)
Supporting information for this article is available on the WWW
under http://dx.doi.org/10.1002/chem.201204558.
plant, Artemisia annua, which is cultivated just for this pur-
pose. An unstable supply, variation in the quality of the har-
vest, and speculation have resulted in heavily fluctuating ar-
temisinin prices. The market is complex because artemisinin
is typically in short supply.[7] Given the price and scarcity of
artemisinin, it may not come as a surprise that up to 40 % of
the ACT drugs sold in Africa are fake and contain either
none or very small amounts of the active ingredient. To
ensure sufficient supplies of high-quality ACTs at prices that
are low enough for all in need to afford them, requires relia-
ble and inexpensive access to artemisinin.
The semisynthesis of artemisinin could supplement cur-
rent drug supplies obtained from isolation and would con-
tribute to a steady supply. Significant efforts have been
channeled into understanding the biosynthesis of artemisi-
nin.[8] The plant produces dihydroartemisinic acid (DHAA),
which gets oxidized by singlet oxygen and further reacts to
give artemisinin,[9] probably without any enzyme involve-
ment.[10] In addition to DHAA, its dehydrogenated precur-
sor, artemisinic acid (AA), can also be found in the plant.
Both AA and DHAA are potential starting materials for
semisyntheses based on a sequence of photooxidation with
singlet oxygen and acid-mediated transformations.[6b, 11] The
amounts of acid precursors and artemisinin present in the
plant vary depending on cultivar, geographic origin, and
time of harvest.[12] Whereas early reports indicated an excess
of artemisinic acid and low levels of artemisinin,[13] im-
proved cultivars can yield more than 1 % artemisinin based
on dry leaf weight.[14] These high yielding varieties of Arte-
misia annua contain minor amounts of artemisinic acid and
support the assumption that dihydroartemisinic acid is the

5450  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eur. J. 2013, 19, 5450 – 5456

direct biosynthetic precursor
of artemisinin.[15] Although a
practical route for the semisyn-
thesis of artemisinin had been
disclosed as early as 1989,[11a]
scale-up of the route that may
well be mimicking the biosyn-
thesis proved extremely chal-
lenging. The photochemical
generation of singlet oxygen
turned out to be difficult to
scale-up in traditional batch
reactors. Nevertheless, the
pharmaceutical company, Scheme 1. Photooxidation products of DHAA.
Sanofi–Aventis, has worked on
a batch synthesis route starting
from artemisinic acid that it procures through a biotechno-
logical process in engineered yeast developed by Amyris.[16]
AA is first reduced into DHAA by chemical means and is
then converted into the corresponding carbonate derivative
to reduce the formation of side products.[17] More recently,
the production of DHAA in yeast has also been reported
and may provide access to this material without the need
for plant extraction.[18]
The key step in the chemical semisynthesis of artemisinin
from DHAA is an ene reaction involving singlet oxygen.
Singlet oxygen can be produced photochemically from trip-
[19]
let oxygen using different sensitizers. The development of
a photochemical batch reactor that ensures uniform irradia-
tion is extremely difficult, if not impossible, as light intensity
decays quickly with increasing distance from the light
source. Continuous-flow chemistry offers a simple solution
to overcome this serious challenge: by wrapping transparent
tubing around a light source short residence times and con-
venient scale-up of photochemical reactions can be ach-
ieved.[20] Under this condition, large specific interfacial area
improves the mass transfer of oxygen from the gas into the
liquid phase.[21] Oxidations with singlet oxygen can be per-
formed efficiently in microstructured systems and small tub-
[22]
ings. We previously communicated the photochemical
continuous synthesis of artemisinin from DHAA, a method
that gave artemisinin in 40 % yield on a 200-g scale per
day.[23]
Herein, we disclose a novel continuous-flow reactor setup
and greatly improved process that produces artemisinin with
high (photo)chemical efficiency. This process was developed
with continuous-flow scale-up in mind. Greatly enhanced se-
lectivity towards artemisinin and high photon efficiency
were realized, to enhance production capacity through a
simplified flow chemistry process.
Results and Discussion
Photooxidation: To design a highly efficient process for the
synthesis of artemisinin, the reaction steps were first opti-
mized separately before the detailed insights helped us to
Chem. Eur. J. 2013, 19, 5450 – 5456  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
FULL PAPER
construct a simplified one-pot continuous process. The first
step in the semisynthesis of artemisinin (1) is the photooxi-
dation of DHAA (2). Singlet oxygen reacts with 2 in an ene
reaction that gives three different hydroperoxides
(Scheme 1). Only hydroperoxide 3 can be converted into ar-
temisinin, whereas the two other endoperoxides, 4 and 5,
contribute to the formation of side products. Hydroperoxide
4 cyclizes to arteannuin H (6) under acidic conditions and
has to be separated from artemisinin at the end of the syn-
thesis. Mercury lamps serve as the predominant light sources
on both industrial[24] as well as laboratory-scale photooxida-
tions. These lamps emit light with a broad range of wave-
lengths, whereas only a fraction of the light is absorbed by
the photosensitizer. Consequently, this process is marked by
relatively poor energy efficiency. A monochromatic light
source that matches the absorption spectrum of the sensitiz-
er is essential for an optimized system. Light-emitting
diodes (LEDs) are monochromatic, are available in various
wavelengths, are highly energy efficient, exhibit a long life-
time, and perform well for photochemical applications.[25] A
high photon flux was achieved by an arrangement of 60
high-power LEDs (72 W electrical-power consumption,
12 W optical output, Fp = 2.5 mmol min 1), emitting at
420 nm, the absorption maximum of tetraphenylporphyrin
(TPP). To irradiate the substrate solution in continuous
flow, the photoreactor was constructed by wrapping tubing
around a transparent glass plate in two layers (7.5 mL
volume). This reactor was mounted at a fixed distance in
front of the LED module. Heat was dissipated from the
LEDs heat sink by a fan, thus minimizing energy input
compared to mercury lamps, for which a powerful cooling
system is essential.
Initially, we explored whether decreased temperatures
during the photooxidation reaction result in improved levels
of selectivity in favor of the desired hydroperoxide, 3. In
contrast to batch photochemistry, synthesis in continuous
flow allows for easy cooling of the photoreactor because
tubing can simply be immersed in a cooled liquid. The small
diameter provides a large interfacial area beneficial for both
fast mass transport of oxygen into the solution and efficient
heat transport between cooling liquid and substrate solution.
www.chemeurj.org 5451
 P. H. Seeberger et al.

Without cooling, the temperature of the photoreactor in-

creased to 46 8C, as measured on the outside of the tubing
when operating the reactor for just four minutes, and to
75 8C when operating for 20 minutes. The photoreactor was
immersed in a cooling/heating bath and the photooxidation
carried out while pumping a 0.5 m solution of DHAA in di-
chloromethane (CH2Cl2) with 1 mm TPP as sensitizer at
1.25 mL min 1 and mixing in oxygen with a T-mixer. The re-
action temperature had a significant effect on the outcome
of the reaction (Table 1). At all temperatures, the starting

Table 1. Temperature effect on the photooxidation step.

Temp. Conv. 3 4 5 Other by-products
[8C] [%] [%] [%] [%] [%] Figure 1. DHAA conversion depends on the sensitizer concentration at a
~ 75[a] 86 62 10 5 24 flow rate of 2 mL min 1 for DCA (&), ZnTPP (*), and TPP (~).
60 99 70 11 5 14
40 99 73 12 4 11
20 99 78 11 4 8 to the pH value and therefore the acid that is necessary for
0 99 81 11 3 5 the subsequent synthesis steps can be added to the reaction
20 98 84 10 3 3
mixture at the beginning. In contrast, the quantum yield of
[a] Reaction carried out without cooling; temperature measured on the TPP is significantly decreased in the protonated form.[27]
outside of the tube. Conv = conversion.

Acid-catalyzed reaction steps: The addition of acid to hy-

material, DHAA, was almost completely consumed, the droperoxide 3 induces a reaction sequence that produces ar-
conversion only dropping slightly with decreasing tempera- temisinin (Scheme 2). Proximal protonation of the hydroper-
ture. At high temperatures, conversion decreases. The de- oxide results in the loss of hydrogen peroxide and the for-
sired hydroperoxide, 3, was obtained as the major product mation of the 5-membered lactone dihydro-epi-deoxyartean-
at all temperatures; however, at lower temperatures, the nuin B (7). Terminal protonation results in a Hock cleavage
ratio was shifted towards the preferred hydroperoxide. The and yields the enol intermediate, 9, which can either react
temperature had an even more pronounced effect on the with triplet oxygen to form a hydroperoxide and ultimately
formation of the other side products. The best yield (84 %) give artemisinin (1), or isomerize to an aldehyde that cycliz-
of desired hydroperoxide 3 was obtained at 20 8C. es to 10. These reactions likely proceed via the ring-expand-
Based on these observations, the photoreactor was cooled ed enol, 8, and the stable enol, 9.[28] Several Lewis and
to 20 8C, working at a substrate flow rate of 2 mL min 1 Brønsted acids had been evaluated previously and trifluoro-
for maximum productivity. The three sensitizers, tetraphe- acetic acid (TFA) was found to perform best to induce
nylporphyrin (TPP, singlet-oxygen quantum yield FD = 0.63 Hock cleavage.[11a, 23] When acetic acid was used instead of
in benzene),[19] zinc tetraphenylporphyrin (ZnTPP, FD = 0.83
in benzene),[19] and 9,10-dicyanoanthracene (DCA, FD = 1.56
in benzene),[26] were evaluated at different concentrations.
These dyes absorb light at 420 nm. Even with relatively low
concentrations of TPP (0.05 mol %), high conversion of
DHAA was achieved (Figure 1). A further increase in con-
centration improved the yield slightly. However, at higher
concentrations, the yield decreased, probably owing to self-
quenching. At 20 8C, selectivity for the desired hydroper-
oxide, 3, was as high as 85 %. The metal complex, ZnTPP,
has a significantly larger quantum yield, but did not perform
well. Most likely, strong photobleaching, as indicated by the
color change after exiting the photoreactor, is responsible
for the poorer performance. The selectivity for the desired
hydroperoxide, 3, dropped to 82 % when ZnTPP was used.
At increased concentration, DCA performed better, as
would be expected based on the higher quantum yield.
Owing to the lower extinction coefficient, a higher concen-
tration of DCA is needed compared to TPP. Levels of selec-
tivity for reactions using either DCA or TPP were similar. Scheme 2. Acid-catalyzed reaction steps yielding artemisinin and side
DCA, which does not contain basic moieties, is insensitive products.

5452 www.chemeurj.org  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eur. J. 2013, 19, 5450 – 5456
Synthesis of Artemisinin
FULL PAPER
TFA, conversion significantly decreased and numerous by-
products appeared. The use of very strong acids, such as sul-
furic acid, led to the conversion of most of the hydroperox-
ide into cyclic aldehyde 10, thus indicating that the tauto-
merisation of the enol intermediate, 9, is too fast to allow
for triplet-oxygen oxidation. Suggested by these observa-
tions, 0.5 equivalents of TFA, based on initial amount of
DHAA, were used for all further reactions.
The influence of solvents on the formation of side prod-
ucts was determined in a batch setup by bubbling oxygen
through a solution of the photooxidation products by using
TFA as acid catalyst. Full conversion of the hydroperoxide,
3, was observed in all solvents. In polar aprotic solvents only
a low yield of artemisinin was obtained, the main products Figure 2. Temperature-dependent acid-catalyzed reaction (20 min) pro-
being the side products, dihydro-epi-deoxyarteannuin B (7) ducing artemisinin (&), dihydro-epi-deoxyarteannuin B (7) (*), side prod-
and 10 (Table 2). The use of solvents of decreased polarity is uct 10 (~), and arteannuin H (6) ( ! ).

Continuous process: After establishing the reac-

Table 2. Yield of artemisinin and dihydro-epi-deoxyarteannuin B based on hydroper-
oxide 3 using TFA as acid catalyst in different solvents. tion parameters of the different steps, a single con-
Solvent Artemisinin Dihydro-epi-deoxyarteannuin B tinuous process was designed. Based on the results
Yield [%] Yield [%] described above, the following setup was expected
acetonitrile 39 36 to result in maximum yield and efficiency: the pho-
dichloromethane 69 17
tooxidation at 20 8C using a sensitizer with high
cyclohexane 76 6
toluene 81 7 quantum yield, such as tetraphenylporphyrin (at
perfluorooctane[a] 40 0 0.75 mm) or dicyanoanthracene (at 2.5 mm), and
benzotrifluoride 78 11 conducted in a nonpolar solvent, such as toluene,
hexafluorobenzene 81 6 benzotrifluoride, or hexafluorobenzene. The acid-
1,3-bis(trifluoromethyl)benzene 82 8
catalyzed step carried out at room temperature
[a] Phase separation occurred. benefits from short residence times and high levels
of selectivity.
beneficial because it leads to the amount of by-products Using DCA as photosensitizer, all reagents, in-
being drastically reduced. Oxygen is very soluble in fluori- cluding the acid, can be added to the initial solution without
nated solvents, which also engender long lifetimes of singlet any loss of efficiency during the photooxidation step. Tol-
oxygen. Whereas the use of fluorinated aromatic solvents uene was chosen as solvent owing to it engendering a higher
prevented the formation of by-products, perfluorooctane selectivity in the process and its lower price. Thus, the con-
was not suitable because the reagents are insoluble in it. Ex- tinuous process requires only one pump, which delivers a
pensive fluorinated solvents have to be recycled, thus ren- toluene solution that contains the starting material, DHAA,
dering toluene the better choice of solvent because it causes the photosensitizer, DCA, as well as the acid catalyst, TFA.
less environmental impact and health concerns. In the con- At least two equivalents of oxygen are added using a T-
text of the flow regime, safety concerns typically associated mixer. The temperature of the solution exiting the photo-
with the use of flammable solvents and oxygen are greatly reactor is adjusted to room temperature and the reaction
reduced because only very small amounts of oxygen are time required to achieve complete conversion determines
present in the microreactor system. the length of tubing. A small and inexpensive system was
To determine the temperature dependence of the acid- constructed (Figure 3). A HPLC pump delivers either sub-
catalyzed reaction cascade, TFA was added to the photooxi- strate solution or pure solvent by controlling a two-way
dation toluene stock solution, the resulting mixture was stir- switch. Oxygen flow is adjusted with a gas-flow controller
red, and oxygen bubbled through the solution for 20 min. and the gas is mixed in using a simple T-mixer. After pas-
The concentration of artemisinin and by-products was deter- sage through the photoreactor, which is cooled to 20 8C,
mined without base quenching. In the absence of oxygen, the solution is reheated in a 10 mL tube reactor and passed
unreacted peroxide 3 is mainly transformed into aldehyde through a second reaction line of 30 mL volume. A back-
by-product 10, which is observed at low temperatures. The pressure regulator (8 bar) at the outlet results in a system
highest yield of artemisinin is obtained at 25 8C (Figure 2). pressure of 10 bar and helps to increase oxygen solubility.
At lower as well as higher temperatures, substantial by- The optimized reactor is fed with a solution of DHAA
product formation and lower yields were observed. At lower (0.5 m), DCA (2.5 mm, 0.5 mol %), and TFA in toluene.
temperatures, the overall artemisinin yield could not be im- Nearly complete conversion was achieved at a substrate
proved by longer reaction times (see the Supporting Infor- flow rate of 1.25 mL min 1 and an oxygen flow of
mation, Table S1). 5 mL min 1. Taking into account the flow rates of both the

Chem. Eur. J. 2013, 19, 5450 – 5456  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim www.chemeurj.org 5453
 P. H. Seeberger et al.

Figure 3. Schematic overview of the continuous-flow reactor.

oxygen gas and the feed solution, residence times can be

calculated. Therefore, the consumption of oxygen during
photoreaction and subsequent thermal oxidation has to be
considered, leading to an apparent decrease in gas-flow rate
over the length of the tubing from 5 mL min 1 (3.2 equiv at
10 bar) to approximately 3.4 mL min 1, after the photoreac-
tor (at 10 bar), and to 1.9 mL min 1, after completion of the
reaction. Assuming average flow rates, the equilibrium resi-
dence time is only 1.5 minutes in the photoreactor and ten
minutes for the thermal oxygenation reaction.
The response surface for two key parameters, equivalents
of acid and reactor temperature after the photoreactor, was
evaluated. For that purpose, both reaction lines were im-
mersed in a water bath and kept at either 0, 15, or 30 8C,
while the feed solution was adjusted to 0.2, 0.6, 1.0, and 2.0
equivalents of TFA. For each parameter combination, 5 mL Figure 4. Dependence of selectivity for artemisinin as a function of tem-
perature of reaction lines 1 + 2 and various amounts of TFA (* 0.2 equiv;
feed solution was injected. At intermediate concentrations
& 0.6 equiv; ~ 1.0 equiv; ! 2 equiv).
of TFA (0.6 and 1.0 equiv) the response surface is very flat
with levels of selectivity in the range of 64 to 68 %. At high
concentrations, selectivity is low, whereas at low TFA con-
centrations, heating is required to drive the reaction to com- tively. The reactor output was collected and, after aqueous
pletion (Figure 4). extraction to remove TFA and removal of solvent, a yellow
A concentration corresponding to 0.5 equivalents of TFA solid was obtained. A crude yield of 65 % of artemisinin at
was selected for further optimization. Considering the 97 % conversion was calculated.
energy required for cooling the photoreactor and the tubing, With these data in hand, the continuous setup was evalu-
the influence of these parameters was investigated (Table 3). ated in terms of (photo)chemical as well as electrical energy
Ideally, the photoreaction would be performed at
room temperature to save energy. However, at Table 3. Temperature effect on the continuous synthesis.
room temperature, the yield of artemisinin in the Photoreactor Tube reactor DHAA Artemisinin Selectivity
reactor output stream is reduced from 66 to 58 %. Temp. [8C] Temp. [8C] Conv. [%] Yield [%] [%]
Cooling the photoreaction mixture results in maxi- 20[a] 0 98 58 60
mum yield. Decreasing the temperature of both re- 20[a] 20 98 57 58
0[a] 0 98 60 62
action lines from 20 8C to 0 8C had only a marginal
0[a] 20 97 62 63
effect. 20[a] 0 98 68 69
For the large scale synthesis of artemisinin, the 20[a] 20 99 69 69
photoreactor was cooled to 20 8C, while the tube 20[b] 20 99 63 64
reactors, 1 and 2, were kept at 10 and 25 8C, respec- [a] In toluene. [b] In dichloromethane.

5454 www.chemeurj.org  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim Chem. Eur. J. 2013, 19, 5450 – 5456
Synthesis of Artemisinin
efficiency. With the standard flow rate of 1.25 mL min 1, a
DHAA conversion of 0.60 mmol min 1 is obtained, which is
24 % of the LEDs photon flow. Even higher levels of
photon efficiency can be realized when a higher feed flow
rate is used, albeit at the expense of complete conversion.
For example, at 2 mL min 1, a DHAA conversion of
0.91 mmol min 1 is obtained. Owing to the limited availabili-
ty of the starting material, incomplete conversion would ne-
cessitate recovery of DHAA from the product stream and,
therefore, we did not pursue this strategy further.
The residence time for the complete process of approxi-
mately 11.5 minutes is very short, but the space–time yield
of the continuous setup is even more impressive. A chemical
yield of 65 % means that the reactor is capable of producing
165 g artemisinin per day. With its volume of 47.5 mL in
total, a space–time yield of 3500 kg m 3 day 1 artemisinin is
calculated.
The energy requirement for the process is an important
aspect, especially because a high-intensity light source is em-
ployed in combination with a low reaction temperature of
20 8C. An overview of the different components power
consumption is shown in Table 4. When active, the chiller
consumed 800 W and during operation the photoreactor
could be kept at 20 8C by using approximately 500 W on
average. Notably, we did not optimize the system concerning
heat transfer. The photoreaction compartment was not iso-
lated and we did not use heat exchangers, conditions that
would be implemented in an industrial setup on larger scale
and could reduce the energy consumption for the cooling
system. However, even the data of the presented setup pro-
vide convincing arguments for the photoreaction to be car-
ried out at 20 8C instead of room temperature. The energy
cost for the chiller is comparably less than the surplus of ad-
ditional artemisinin owing to better selectivity.
For purification, the majority of the dye was removed by
dissolving the crude product in acetonitrile before passing
the suspension through a PTFE syringe filter, owing to the
low solubility of DCA. The crude product was recrystallized
from cyclohexane/ethanol (9:1 v/v) and off-white needles of
artemisinin with some remaining DCA were obtained (57 %
yield upon isolation). A second recrystallization from cyclo-
hexane/ethanol (9:1 v/v), yielded pure white needles (46 %
yield based on initial DHAA) as assessed by HPLC by
using UV absorbance, ELSD, and MS detection (see the
Supporting Information, Figure S7, S8, S9).[29] Owing to the
Table 4. Energy consumption and cost of the reactor components.
Component Power Electrical energy
consumption per kg 1
[W] ACHTUNGRE[kWh]
LED lamp + control unit 100 14.5
+ power supply
fan for lamp 4 0.58
pump 6 0.87
chiller 500 72.7
Total 610 88.7
Chem. Eur. J. 2013, 19, 5450 – 5456  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
FULL PAPER
low extinction coefficient of artemisinin compared to
common impurities, UV absorbance at 210 nm provides an
efficient means for the detection of trace impurities.[29] Four
minor peaks account for just 1.3 area % of the UV absorb-
ance signals in total.
Plant extract as starting material: Both AA and DHAA are
present in the plant, but are currently discarded in large
quantities. When artemisinin is extracted from Artemisia
annua, the final purification is a crystallization process leav-
ing mother liquor behind that is treated as waste. The analy-
sis of mother liquor obtained from an extractor was found
to contain 2.0 % artemisinic and 8.2 % dihydroartemisinic
acid. Simple basic extraction of the mother liquor produced
a material that contained 42 % DHAA based on 93 % re-
covery.[30] A solution of this crude extract in toluene in the
presence of TFA (0.25 m) and DCA (2.5 mm) was subjected
to the continuous-flow synthesis process. Best results were
obtained when the reactor eluent was quenched with the
more basic potassium carbonate rather than sodium bicar-
bonate and artemisinin was obtained in 57 % crude yield
based on DHAA content (see the Supporting Information,
Figure S4). Artemisinin was recrystallized in high purity
from cyclohexane/ethanol (9:1 v/v) with 73 % recovery.
Plant waste may thus serve as feed stock by providing
DHAA for the semisynthesis of artemisinin.
Conclusion
A careful assessment of the reaction parameters that influ-
ence the outcome of the continuous-flow semisynthesis of
the antimalarial drug, artemisinin, from dihydroartemisinic
acid resulted in a greatly simplified process and a signifi-
cantly improved yield. Now, the continuous-flow reaction re-
quires just one pump and an initial supply of oxygen. The
compact setup relies on a LED lamp with sensitizer-tailored
emission wavelength, resulting in a photon efficiency of
24 %.
Based on the in depth understanding of the synthesis, a
larger flow reactor that will be able to produce one metric
ton of artemisinin per year is currently being constructed.
The energy efficiency can be improved through better isola-
tion of the photoreactor, the use of heat exchangers, and a
different chiller. The fact that dihydroartemisinic acid can
be obtained from the waste of current extraction
processes provides an alternative to starting materi-
al derived from fermentation in engineered yeast,
Cost per kg 1 thus making the process even more attractive.
(0.2 E/kWh)
[E]
2.9
Experimental Section
0.1
0.2
Power consumption was measured with an electricity meter
14.5
(Voltcraft, Energy Check 3000). A HPLC pump (Knauer,
Smartline pump 100) was used to deliver either substrate solu-
17.7
tion or pure solvent by controlling a two-way switch. Pressure
www.chemeurj.org 5455

was measured at the exit of the pump head with a built-in sensor. In an
ETFE T-mixer (IDEX Health and Science) substrate solution was mixed
with oxygen, which was delivered through a check valve from an oxygen
gas tank. Gas pressure was regulated to 15 bar and the flow adjusted to
5 mL min 1 with a gas-flow controller (Influx, SV1B5-AI05). This solu-
tion was then pumped through the photoreactor, consisting of fluorinated
ethylene–propylene copolymer (FEP) tubing (IDEX Health and Science,
natural color, 1.57 mm outer diameter, 0.76 mm inner diameter) of
7.5 mL volume wrapped around a glass plate in two layers (7  9 cm2).
The plate was immersed in a glycol/water (3:2 v/v) bath, cooled to
20 8C with an immersion cooler (Huber, TC100E-F-NR). The LED
module (OSA Opto Light, OLM-018 B, 420 nm emission wavelength)
was mounted in front of this plate at a distance of 3 cm. After passage
through the photoreactor, the solution was reheated to 10 8C in a reaction
line of 10 mL volume (1.57 mm outer diameter, 0.76 mm inner diameter)
and a second reaction line of 30 mL volume (3.18 mm outer diameter,
1.57 mm inner diameter) kept at room temperature. A back-pressure reg-
ulator of 8 bar was employed at the end of the system and the output
stream was collected in an Erlenmeyer flask. 250 mL of feed (0.5 m
DHAA: 29.5 g, 0.25 m TFA: 4.8 mL, and 2.5 mm DCA: 143 mg in tol-
uene) were injected. The collected solution was extracted twice with
aqueous sat. NaHCO3 to remove TFA, yielding, after removal of toluene,
a yellow solid. The majority of dye was removed by solubilizing the
crude in acetonitrile and filtering the suspension through a PTFE syringe
filter with 0.45 mm cutoff. Recrystallization from cyclohexane/ethanol
(9:1 v/v, 150 mL) yielded off-white needles (16.52 g, 47 % yield upon iso-
lation). The mother liquor was recrystallized from cyclohexane (50 mL),
yielding slightly yellow crystals, which were washed with cyclohexane and
dried (3.60 g, consisting of artemisinin (96 % purity as determined using
NMR spectroscopy, 3.45 g), 10 % yield upon isolation). Both artemisinin
batches were combined and recrystallized from cyclohexane/ethanol (9:1
v/v, 150 mL), yielding artemisinin as white needles (16.08 g, 46 % yield
upon isolation based on initial DHAA). M.p. 154–155 8C; [a]20 D = + 68
(c = 1.00 in CHCl3); 1H NMR (400 MHz, CDCl3, 25 8C, TMS): d = 5.85 (s,
1 H), 3.39 (qd, 3JACHTUNGRE(H,H) = 7.3, 5.4 Hz, 1 H), 2.48–2.35 (m, 1 H), 2.09–1.94
(m, 2 H), 1.92–1.84 (m, 1 H), 1.81–1.71 (m, 2 H), 1.50–1.33 (m, 3 H), 1.44
(s, 3 H), 1.20 (d, 3JACHTUNGRE(H,H) = 7.3 Hz, 3 H), 1.12–1.02 (m, 2 H), 1.00 ppm (d,
3
JACHTUNGRE(H,H) = 6.0 Hz, 3 H); 13C NMR (101 MHz, CDCl3, 25 8C,TMS): d =
172.0, 105.4, 93.7, 79.5, 50.1, 45.0, 37.5, 35.9, 33.6, 32.9, 25.2, 24.9, 23.4,
19.8, 12.6 ppm; HRMS (ESI): m/z calcd for C15H22O5 + Na + : 305.1359
[M+Na + ]; found: 305.1395.
Acknowledgements
We gratefully acknowledge financial support from the Max-Planck Soci-
ety. Dr. Siems (AnalytiCon Discovery, Potsdam, Germany) kindly ana-
lyzed the mother liquor from the extraction of Artemisia annua that was
provided by Bionexx/Innovexx, (Antananarivo, Madagascar).
[1] World Malaria Report, World Health Organization, Geneva, 2011.
[2] T. K. Mutabingwa, Acta Trop. 2005, 95, 305 – 315.
[3] Y. Tu, Nat. Med. 2011, 17, 1217 – 1220.
[4] G. H. Posner, S. B. Park, L. Gonzalez, D. Wang, J. N. Cumming, D.
Klinedinst, T. A. Shapiro, M. D. Bachi, J. Am. Chem. Soc. 1996, 118,
3537 – 3538.
[5] a) Y. Li, Y. L. Wu, Curr. Med. Chem. 2003, 10, 2197 – 2230; b) R. K.
Haynes, Curr. Top. Med. Chem. 2006, 6, 509 – 537; c) A. R. Kreeft-
meijer-Vegter, P. J. van Genderen, L. G. Visser, W. F.-W Bierman, J.
Clerinx, C. K.-W van Veldhuizen, P. J. de Vries, Malar. J. 2012, 11,
102.
[6] a) G. Schmid, W. Hofheinz, J. Am. Chem. Soc. 1983, 105, 624 – 625;
b) H. J. Liu, W. L. Yeh, S. Y. Chew, Tetrahedron Lett. 1993, 34,
4435 – 4438; c) J. S. Yadav, B. Thirupathaiah, P. Srihari, Tetrahedron
2010, 66, 2005 – 2009; d) C. Zhu, S. P. Cook, J. Am. Chem. Soc. 2012,
5456 www.chemeurj.org  2013 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim
P. H. Seeberger et al.
134, 13577 – 13579; e) X. X. Xu, J. Zhu, D. Z. Huang, W. S. Zhou,
Tetrahedron 1986, 42, 819 – 828.
[7] a) R. van Noorden, Nature 2010, 466, 672 – 673; b) A. Maxmen, Nat.
Med. 2012, 18, 634 – 635.
[8] a) G. D. Brown, Molecules 2010, 15, 7603 – 7698; b) R. K. Haynes,
S. C. Vonwiller, Acc. Chem. Res. 1997, 30, 73 – 79.
[9] L. K. Sy, G. D. Brown, Tetrahedron 2002, 58, 897 – 908.
[10] P. S. Covello, Phytochemistry 2008, 69, 2881 – 2885.
[11] a) R. J. Roth, N. Acton, J. Nat. Prod. 1989, 52, 1183 – 1185; b) R. K.
Haynes, S. C. Vonwiller, J. Chem. Soc., Chem. Commun. 1990, 451 –
453.
[12] a) T. E. Wallaart, N. Pras, A. C. Beekman, W. J. Quax, Planta Med.
2000, 66, 57 – 62; b) H. J. Woerdenbag, N. Pras, N. G. Chan, B. T.
Bang, R. Bos, W. Vanuden, P. V. Y, N. V. Boi, S. Batterman, C. B.
Lugt, Planta Med. 1994, 60, 272 – 275.
[13] M. Jung, H. N. Elsohly, E. M. Croom, A. T. McPhail, D. R. McPhail,
J. Org. Chem. 1986, 51, 5417 – 5419.
[14] N. Delabays, X. Simonnet, M. Gaudin, Curr. Med. Chem. 2001, 8,
1795 – 1801.
[15] a) J. F. S. Ferreira, D. L. Luthria, J. Agric. Food Chem. 2010, 58,
1691 – 1698; b) W. J. M. Lommen, S. Elzinga, F. W. A. Verstappen,
H. J. Bouwmeester, Planta Med. 2007, 73, 1133 – 1139.
[16] D. K. Ro, E. M. Paradise, M. Ouellet, K. J. Fisher, K. L. Newman,
J. M. Ndungu, K. A. Ho, R. A. Eachus, T. S. Ham, J. Kirby, M. C. Y.
Chang, S. T. Withers, Y. Shiba, R. Sarpong, J. D. Keasling, Nature
2006, 440, 940 – 943.
[17] a) J. Dhainaut, A. Dlubala, R. Guevel, A. Medard, G. Oddon, N.
Raymond, J. Turconi, (Sanofi – Aventis, Fr.), WO 2011/026865, 2011;
b) V. Kraft, G. Kretzschmar, K. Rossen, (Sanofi – Aventis, Fr.), WO
2011/030223, 2011.
[18] a) Y. Zhang, K. H. Teoh, D. W. Reed, L. Maes, A. Goossens, D. J. H.
Olson, A. R. S. Ross, P. S. Covello, J. Biol. Chem. 2008, 283, 21501 –
21508; b) P. J. Westfall, D. J. Pitera, J. R. Lenihan, D. Eng, F. X.
Woolard, R. Regentin, T. Horning, H. Tsuruta, D. J. Melis, A.
Owens, S. Fickes, D. Diola, K. R. Benjamin, J. D. Keasling, M. D.
Leavell, D. J. McPhee, N. S. Renninger, J. D. Newman, C. J. Paddon,
Proc. Natl. Acad. Sci. USA 2012, 109, E111 – E118.
[19] M. C. DeRosa, R. J. Crutchley, Coord. Chem. Rev. 2002, 233, 351 –
371.
[20] B. D. A. Hook, W. Dohle, P. R. Hirst, M. Pickworth, M. B. Berry,
K. I. Booker-Milburn, J. Org. Chem. 2005, 70, 7558 – 7564.
[21] K. Jhnisch, V. Hessel, H. Lowe, M. Baerns, Angew. Chem. 2004,
116, 410 – 451; Angew. Chem. Int. Ed. 2004, 43, 406 – 446.
[22] a) R. C. R. Wootton, R. Fortt, A. J. de Mello, Org. Process Res. Dev.
2002, 6, 187 – 189; b) F. Lvesque, P. H. Seeberger, Org. Lett. 2011,
13, 5008 – 5011; c) R. A. Bourne, X. Han, M. Poliakoff, M. W.
George, Angew. Chem. 2009, 121, 5426 – 5429; Angew. Chem. Int.
Ed. 2009, 48, 5322 – 5325.
[23] F. Lvesque, P. H. Seeberger, Angew. Chem. Int. Ed. 2012, 51, 1706 –
1709.
[24] W. Pickenhagen, D. Schatkowski, (Dragoco Gerberding & Co. AG,
Germany), US 005892059 A, 1999.
[25] F. R. Bou-Hamdan, P. H. Seeberger, Chem. Sci. 2012, 3, 1612 – 1616.
[26] R. C. Kanner, C. S. Foote, J. Am. Chem. Soc. 1992, 114, 678 – 681.
[27] V. S. Chirvony, A. van Hoek, V. A. Galievsky, I. V. Sazanovich, T. J.
Schaafsma, D. Holten, J. Phys. Chem. B 2000, 104, 9909 – 9917.
[28] S. C. Vonwiller, J. A. Warner, S. T. Mann, R. K. Haynes, J. Am.
Chem. Soc. 1995, 117, 11098 – 11105.
[29] R. W. Stringham, M. Pennell, W. Cabri, G. Carzana, F. Giorgi, S.
Lalli, G. Marazzi, M. Torri, J. Chromatogr. A 2011, 1218, 6838 – 6842.
[30] a) T. E. Wallaart, W. van Uden, H. G. M. Lubberink, H. J. Woerden-
bag, N. Pras, W. J. Quax, J. Nat. Prod. 1999, 62, 430 – 433; b) R. K.
Haynes, S. C. Vonwiller, Trans. R. Soc. Trop. Med. Hyg. 1994, 88,
23 – 26.
Received: December 21, 2012
Revised: March 2, 2013
Published online: March 20, 2013
Chem. Eur. J. 2013, 19, 5450 – 5456