Synthesis of polyimides in supercritical carbon dioxide�
Ernest E. Said-Galiyev a,*, Yakov S. Vygodskii a, Lev N. Nikitin a,Rostislav A. Vinokur a, Marat O. Gallyamov b, Inna V. Pototskaya c,Vyacheslav V. Kireev c, Alexei R. Khokhlov a,b, Kjeld Schaumburg d
a A.N. Nesmeyanov Institute of Organoelement Compounds, Russian Academy of Sciences, Vavilova Street, 28, Moscow 119991, Russiab Physics Department of Lomonosov State University, Vorobievy gory, Moscow 119992, Russia
c D.I. Mendeleev University of Chemical Technology of Russia, Moscow, Russiad Centre for Interdisciplinary Studies of Molecular Interactions, Department of Chemistry, University of Copenhagen, Universitetsparken
5, DK-2100 Copenhagen, Denmark
Received 15 September 2001; received in revised form 30 August 2002; accepted 5 September 2002
Abstract
One-step polycyclization reaction in supercritical carbon dioxide has been carried out for the first time and a thereby
set of polyimides has been synthesized. Diphenyl-2,2-hexafluoropropane-3,3?,4,4?-tetracarboxylic acid dianhydride
(dianhydride 6F), 4,4?-diaminodiphenyl-2,2-hexafluoropropane (diamine 6F) and 9,9-bis(4-aminophenyl)fluorene were
used as monomers. Maximum inherent viscosity of the synthesized polymers is 0.56 dl/g. It is supposed that carbon
dioxide act as an acidic catalyst for the reaction in the presence of water admixture.
# 2003 Elsevier B.V. All rights reserved.
Keywords: Supercritical carbon dioxide; Polycyclization reaction; Synthesis; Polyimides
1. Introduction
Polymer synthesis in supercritical (sc) solvents is
a rapidly developing branch of the environment-
friendly chemistry [1]. Many types of radical
polymerization in sc solvents have already been
studied. The properties of the polymers obtained
by such reactions have not yet been studied
extensively. It was however shown that in most
cases without catalysts sc CO2 is an absolutely
neutral solvent. It was also shown that the sc
solvents can be an effective substitute for freons
and other undesirable halogen-containing organic
solvents [2].In contrast to the polymerization reactions, the
polycondensation reactions are by far not so well
studied [3�/9]. It can be due to a more complex
polycondensation technique in sc CO2.
Recently we reported the first synthesis of
polyimides (PI) in sc CO2 [10�/12]. Now some
new results and more detailed description of the
synthesis are presented. It is well known that
carboxylic acids, including acetic and benzoic
acid are effective catalyst for PI synthesis by the
�PII of original article: S 0 8 9 6 - 8 4 4 6 ( 0 2 ) 0 0 2 1 0 - 3
* Corresponding author. Tel.: �/7-095-135-05-22; fax: �/7-
095-135-50-85.
E-mail address: [email protected] (E.E. Said-Galiyev).
J. of Supercritical Fluids 27 (2003) 121�/130
www.elsevier.com/locate/supflu
0896-8446/03/$ - see front matter # 2003 Elsevier B.V. All rights reserved.
doi:10.1016/S0896-8446(03)00146-3
reaction of tetracarboxylic acid dianhydrides anddiamines [13]. Since CO2 in the presence of water
admixture is acting as a weak acid, we suppose
that sc CO2 may act not only as a reaction
medium, but also as a catalyst for this reaction.
2. Experimental
2.1. Experimental materials
The choice of the monomers for a polyconden-
sation reaction was based on excellent solubility of
fluorinated PI in many organic solvents and a
good solubility of fluorinated polymers in sc CO2.
Diphenyl-2,2-hexafluoropropane-3,3?,4,4?-tetra-carboxylic acid dianhydride (dianhydride 6F) and
4,4?-Diaminodiphenyl-2,2-hexafluoropropane
(Diamine 6F) were received from ‘Aldrich’ and
have been purified by high vacuum sublimation.
Dianhydride A was purchased from ‘General
Electric’ (m.p., 186 8C). 9,9-bis(aminophe-
nyl)fluorene (Diamine AF) (m.p., 236�/237 8C)
was synthesized in our laboratory and purifiedby high vacuum sublimation [13]. Two kinds of
carbon dioxide was used: normal purity (�/99.0%)
and high purity (�/99.997%).
2.2. Synthesis technique
PI were synthesized by high temperature one-
step polycyclization reaction of the corresponding
diamines and dianhydrides.
The synthesis was performed in a high-pressure
cell with inner volume 10 cm3 described in detail
elsewhere [4]. A standard set-up was used to obtain
sc conditions in the cell. The reagents were loaded
into the cell (dianhydride 6F, 0.3 g, 6.75�/10�4
M, diamine AF 0.188 g, 5.4�/10�4 M and/or
diamine 6F 4.5�/10�2 g, 1.35�/10�4 M/0.2254 g,
6.75�/10�4 M with molar ratio 1:0.8:0.2/1:0:1.
The cell was then purged with CO2, sealed and
heated up to the temperature of 180 8C. We
intended to elucidate whether CO2 plays a specific
role in the polycyclization reaction. The role of sc
CO2 in reactions in a melt is very well known. Sc
CO2 reduces the melt viscosity that promotes a
better elimination of low molecular mass reaction
products and increases the polymer MM [2]. It was
not possible to elucidate the specific role of CO2 in
a polycondensation reaction at the temperature
higher than the m.p. of monomers.
It is well known that sc CO2 decreases Tg of
polymers owing to its plasticizing action [14].
Therefore the m.p. of the monomers could be
expected slightly to decrease.
When dianhydride A was used later as dianhy-
dride component the reaction temperature ap-
proached probably close to a melting
temperature of the reaction mass.
A syringe press provided by ‘High Pressure
Equipment’ was used to pump the liquid CO2
into the cell up to a pressure of 32.5 MPa where-
upon a magnetic stirrer was switched on. The
reaction time was varied from 3 to 6.5 h and it was
chosen based on known data about synthesis of
high MM PI under such conditions. Both batch
and flow modes of operation (persistent flow of sc
CO2) were used. CO2 flow rate was 15�/20 cm3/
min. The temperature was controlled with an
accuracy of 9/0.5 degrees, and the pressure within
9/1 bar.The cell was cooled after of run, the stirrer was
switched off and the cell was decompressed. The
Erratum 123
polymer was further reprecipitated from acetoneor 1-methyl-2-pyrrolidinone solution into excess
water and then the sample was dried overnight in
air and in vacuum at 50 8C for 1 h.
2.3. Methods of analysis
FT-IR spectral measurements were performed
using Nicolet ‘Magna 750’ spectrometer. The
specimen used for spectroscopy were KBr tablets
with added polymer powder.
Molecular weight analysis was performed with
GPC technique in THF solvent (flow rate is 1 ml/
min) using ‘Waters’ chromatograph consisting of
M600 delivery system, M 484 UV-VIS detector
(l�/260 nm) and U-Styrogel linear columns. The
chromatograph was calibrated on PS standards.
MAXIMA software was used.
Table 1
Run conditions
No. Dianhydride Diamine Molar ratio Temperature, 8C Pressure, bars Time Medium Regime
1 6F 6F�/AF 1:0.2:0.8 180 325 3 sc CO2 Batch
2 6F 6F�/AF 1:0.2:0.8 130 325 3 sc CO2 Batch
3 6F 6F�/AF 1:0.2:0.8 180 1 3 ?r Batch
3a 6F 6F�/AF 1:0.2:0.8 180 1 3 Ar Flow
3b 6F 6F�/AF 1:0.2:0.8 180 325 3 sc CO2 Flow
4 6F 6F�/AF 1:0.2:0.8 180 325 4 sc CO2 Flow
5 6F 6F�/AF 1:0.2:0.8 180 325 6 sc CO2 Batch
6 6F 6F�/AF 1:0.2:0.8 180 325 6 sc CO2 Batch
7 6F 6F�/AF 1:0.2:0.8 180 325 6.5 sc CO2 Flow
8 6F 6F 1:1 180 325 6.5 sc CO2 Flow
9 6F 6F�/AF 1:0.2:0.8 180 325 6.3 sc CO2a Flow
10 6F 6F 1:1 180 325 6.0 sc CO2b Flow
11 A 6F�/AF 1:0.2:0.8 180 325 6.0 sc CO2 Flow
12 A AF 1:1 180 325 4.0 sc CO2 Flow
a High-purity CO2 (�/99.997%).b High-purity CO2 (�/99.997%), special addition of water equal in quantity to that introduced into the cell with normal purity CO2
(99.0%) in batch mode of operation.
Table 2
Polymer characteristics
Appearance Yield, % Inherent viscosity, dl/g Imide cycle bands,
cm�1
Anhydride cycle bands,
cm�1
721 1380 1730 1780 1860
1 Light-yellow powder 93.7 0.14 m s s m w
2 Light-yellow powder 91.0 0.12 m s s w m
3 Yellow powder 94.6 0.19 w w m s s
4 Brown powder 81 0.26 s s s m m
5 Dark-yellow powder and brown crystal 100.0 0.12 m s �/ w
6 Dark-yellow powder and brown crystal 91.5 0.19 m s s w w
7 Dark-yellow powder and brown crystal 88.0 0.32 m s s w w
8 Light-yellow powder 80.0 0.43 m s s w �/
9 Light-yellow uniform powder 88.0 0.14 w w w s s
10 Grey powder 88.3 0.13 m s s w �/
11 Light-yellow powder 90.0 0.56 m s s w �/
12 Pale-yellow powder 93.0 0.40 m s s w �/
s, strong; m, medium; w, weak; �/, absent.
Erratum124
A solubility of the monomers was determind
visually by observation of a sample powder
particles behavior in a little glass located in the
spectroscopic cuvette filled with sc CO2. The
samples also were weighed before and after
exposition in sc CO2. A change of crystals form
without a weight loss was attributed to a slight
swelling.
Inherent viscosity (/h?inh) was measured at con-
centration of 0.5 g/dl in 1-methyl-2-pyrrolidinone
at 25 8C, using a 10 ml Ubbelohde viscometer.
These data permit a qualitative evaluation of the
Polyimide MM value because the intrinsic viscos-
ity and Polyimide MM are related via Mark�/
Kuhn Howink equation: [h ]tsolvent�/KMa , where
K and a are constants depending on the polymer
and the solvent. The experimental conditions are
shown in Table 1.
3. Results and discussion
Solubility of reagents is one of the crucial issues
for the polycondensation reaction. It is especially
important in the case of CO2 because only a few
classes of polymers have a considerable solubility
in sc CO2. The best known classes are fluorine-
containing polymers (polyfluoroalkyl acrylates,
fluorine-substituted polyethers, some of the poly-
perfluoroolefines and also their block and random
copolymers) and several silicon polymers (poly-
dimethylsiloxanes) [1].
Data on a new class of CO2-philic polymers
have been published recently. The polymers are
block copolymers of polyalkylene carbonate and
polyethylene oxide [15].
In our case the solubility tests showed that only
diamine 6F is soluble in sc CO2 among the
Fig. 1. FT-IR spectrum of polyimide (/h?inh 0.32 dl/g). Run conditions: dianhydride 6F and diamines mixture (6F�/AF) with molar ratio
(0.2:0.8), T�/180 8C, P�/325 bar, t�/6.5 h.
Erratum 125
reaction components under the run conditions.
Dianhydrides 6F and A exhibit a small swelling
degree, diamine AF and the polymers are insoluble
in sc CO2. Thus we suppose that the reaction
occurs in a heterogeneous medium. Fluorinated
polymers show a good solubility in many low-
temperature solvents: acetone, chloroform, THF,
etc. Such behavior is typical for aromatic PI
synthesized in high boiling solvents from dianhy-
drides and diamines containing bridging fluori-
nated groups [3]. The properties of the polymers
are shown in Table 2.
The polymer yield was practically quantitative
despite the heterogeneous reaction conditions.
Polymer viscosities in batch mode are rather low
(runs 1, 2, 5 and 6). Using flow mode and similar
other reaction parameters allows viscosity increase
approximately by a factor of 1.7 (up to 0.32 dl/g,
runs 6 and 7). Polymers synthesized in flow mode
show a high intensity of the imide cycle bands
(721, 1380 and 1730 cm�1). At the same time the
bands of anhydride cycle at 1860 cm�1, and those
of amide bond and 1654 and 3400 cm�1 are either
weak or absent at all (Fig. 1 and Tables 1 and 2;
runs 4, 7, 8, 11 and 12).
Polymers synthesized in batch mode show a low
intensity of anhydride bands, too. Therefore batch
mode (as well as flow mode) allows to reach a high
conversion monomers. However, in this case
oligomers are mostly produced. They should
have sufficiently great amount of end groups.
One of the reasons of why IR spectra of the
polymers synthesized in batch mode show a low
intensity of bands corresponding to anhydridegroups is both steps of polycondensation: forma-
tion of polyamide acid and its cyclization occur
almost simultaneously and then are followed by
chain growth. At the same time according to
Carothers: degree of polymerization DP�/1/1�/
p , where p is conversion. If p�/0.95 (it is close
to sensitivity limit of routine IR-spectroscopy
method) DP�/20 (only). On the other hand wehave the sample with h?inh 0.32 dl/g and
Mn(GPC)�/4700. MM of the elemental unit is
734. Simple calculations give p�/0.92. But never-
theless the anhydride groups in FT-IR spectrum of
the sample are practically absent (Fig. 1).
Equilibrium concentration of water in batch
mode is rapidly reached and chain growth is
ended. Flow mode is specially used to removethe excess of water. This allows to shift the
equilibrium towards formation of polymers having
higher molecular mass (Table 2, runs 6 and 7).
Polycyclization reaction takes easily place under
a wide change of run conditions. Thus polymers
were synthesized even in argon atmosphere (runs 3
and 3a), but they had low h?inh:/Viscosity values equal 0.14 dl/g (run 1) and 0.19
dl/g (run 3) indicate a low molecular mass poly-
mer. However, viscosity of polymers synthesized in
argon was somewhat greater than of those synthe-
sized in CO2. This might lead to the conclusion
that CO2 does not play any role in batch mode
synthesis. One can suppose the humid CO2 in the
batch conditions brings more quantities of the
water in the system compared with Ar and therebyit promotes an achievement of an equilibrium in
condensation�/hydrolysis reaction for shorter time,
that influences the chain length. The imide bands
of the polymer in run 3 have medium intensity
(very intensive in run 1), anhydride bands in both
runs are very low and amide bands in run 3 have
more intensity higher than medium (medium one
in run 1). Thus the reaction in sc CO2 in batchmode promotes a higher conversion of polyamic
acid into polyimide.
The runs in argon and CO2 in flow mode were
performed (Table 1 and Table 3, runs 3a and 3b)
to determine the role of CO2 in polycyclization.
Since the viscosities of the polymers in both cases
are similar, we concluded that there is no con-
Table 3
The properties of polyimide synthesized in Ar (3a) and CO2
(3b) gas flow, 180 8C, 3 h
Run numbers 3a 3b
Physical properties Yellow powder Canary powder
Yield, % 60 74
Inherent viscosity, dl/g 0.17 0.20
Imide cycle bands, cm�1 721m, 1380m,
1730m
721m, 1380s 1730s
Anhydride cycle bands,
cm�1
1860w, 1780w 1860 is absent,
1780w
Amide group bands 1654s, 3400s 1654w, 3400w
s, strong; m, medium; w, weak.
Erratum126
siderable equilibrium shift in the polycyclization
reaction under these conditions (perhaps a rate of
water removal is not enough). Comparison of the
spectral data (Fig. 2 and Fig. 3, Table 3, runs 3a
and 3b) shows the bands corresponding to the
anhydride end groups are practically absent for the
polymer synthesized in sc CO2 (very low for
polymer synthesized in Ar) and those correspond-
ing to the amide bond (1654 and 3400 cm�1) have
a relatively low intensity, whereas they are very
intensive for polymer synthesized in argon. Imide
cycle bands (721, 1370 and 1730 cm�1) are more
intensive for the CO2-sample as compared with Ar
one. Polyimide is slowly formed in Ar atmosphere
as well (Table 3). Thus a moderate monomers
conversion into PI occurs also in Ar atmosphere
however an essential part of the polymer is present
as polyamic acid structure. Synthesis in sc CO2
leads to higher polyimide yield. These data give an
evidence of accelerating catalytic effect of CO2 in
both reaction steps.
If there is a catalytic effect in sc CO2 the same
one as we supposed must be observed in a CO2
flow as compared with Ar one at the pressure of 1
bar. We have carried out such runs. The results are
given in Table 4.
It is seen from the table CO2 provides for a
higher polymer viscosity and yield. These data also
point to the catalytic action of CO2 in PI synthesis.
It is well known that carboxylic acids catalyze
the polyimidation reaction [13]. We used high
purity (�/99.997%, H2O�/0.001%) to prove a
participation of water in catalysis (Table 2, run
9). All the other conditions were the same as in the
run 8. Polymer with low viscosity was obtained.
The result was reproduced several times and the
polymer inherent viscosities were in the range of
0.14�/0.16 dl/g. Therefore we can suggest that sc
CO2 can act as an acidic catalyst in the presence of
a water admixture in the system.On the other hand, an attempt to create
catalytic conditions by adding water to the reac-
tion cell while using high-purity CO2 did not lead
to synthesis of high MM polymer (Table 2, run
10). It may be due to excessive quantity of
added water. There are no exact data on the
minimum concentration of water necessary for
catalysis.
Fig. 2. FT-IR spectrum of polyimide (/h?inh 0.17 dl/g). Monomers: dianhydride 6F and diamine mixture (6F�/AF) with molar ratio
(0.2:0.8) T�/180 8C, P�/1 bar (Ar flow), t�/3 h.
Erratum 127
Polymer viscosity values in runs 6 and 7 lead to
the conclusion that the synthesis mode plays
nevertheless a certain role in sc conditions namely
the flow mode of operation gives rise to higher
MM.
As it can be seen from Table 2, decrease of the
temperature in the run by 50 8C in batch mode
does not influence the MM (runs 1 and 2).
The increase of the reaction time from 3 to 6 h in
batch mode leads to a certain increase of the
inherent viscosity (runs 1 and 6). Therefore an
equilibrium is reached under these conditions in
more than 3 h. Increase of the reaction time in flow
mode leads to an increase of inherent viscosity too
(runs 4 and 7). If the reaction is carried out in flow
mode from reaction start, then monomers dis-
solved in CO2 can be carried away by the gas flow
and stoichiometry can be changed.
On the other hand as it was noted in [2], in case
of polycarbonate synthesis from 4,4-bis(hydroxy
Fig. 3. FT-IR spectrum of polyimide with h?inh 0.20 dl/g. Monomers: dianhydride 6F and diamine mixture (6F�/AF) with molar ratio
(0.2:0.8), 180 8C, 325 bar (in CO2 flow), 3 h.
Table 4
Dependence of polymer inherent viscosity and yield on gas
nature
Gas Inherent viscosity, dl/g Yield, %
Argon 0.43 84
Carbon dioxide 0.51 89
The conditions: 180 8C, 1 bar, m -cresol, 3 h.
Erratum128
phenyl)-2,2-propane and diphenylcarbonate, theoligomer precipitates from the solution in sc CO2
already after 1�/2 condensation acts. Therefore the
initial phase (1.5 h) of all the flow mode runs with
monomers soluble in sc CO2 was performed in
batch mode. Molecular mass of a sample with
inherent viscosity 0.32 dl/g was determined with
gel-permeation chromatographic technique: Mw�/
12 400, Mn�/4700, Mw/Mn�/2.6. Polymers withsimilar MM were synthesized with heterogeneous
polymerization: Polycarbonate (Mw�/4500�/
27 000, Mn�/2200�/11 000) and PET (M�/3000�/
6300) in sc CO2 [2].
Such moderate MM are most likely due to the
heterophase reactions conditions. It is known that
the reactions in solid phase develop slower than in
liquid phase. Heterophase conditions prevent theremoval of low molecular weight reaction products
and shift the equilibrium towards lower monomers
conversion, thus preventing the formation of high
MM polymers.
Nevertheless we succeeded in subsequent runs in
the synthesis of PI with h?inh up to 0.56 dl/g under
heterogeneous conditions. Therefore heteroge-
neous conditions in given system are not theprincipal limiting factor for reaching high MM.
On the other hand a higher degree of homogeniza-
tion facilitates the increase of polyimide MM.
(Table 1 and Table 2, runs 8, 12 and 11. In these
runs soluble diamine 6F was used. Homogeniza-
tion can also increase due to dianhydride A
melting because of depression of its melting point
in sc CO2.)Run 11 extends the set of potential monomers.
Run 12 shows that polymers with relatively high
molecular mass can be synthesized using mono-
mers insoluble in sc CO2. This opens wide
perspectives for carbon dioxide substitution of
toxic solvents in polyimide production with envir-
onment-friendly.
4. Conclusions
1) PI have been synthesized by one-step poly-
cyclization in sc carbon dioxide for the first
time.
2) A solubility of monomers is not a limitingcriterion for achievement of high molecular
weight of PI.
3) The flow mode of synthesis leads to an
increase of polyimide molecular weight.
4) The hypothesis about catalytic action of CO2
in synthesis of PI in the presence of water
admixture is suggested. Some confirming ex-
perimental data have been obtained.
Acknowledgements
This work was supported by Russian Founda-
tion for Basic Research, projects 02-03-32089 and
No. 01-03-32766a, by the Danish Research Coun-
cil THOR program FUCOMA. and by NATO
Programme ‘Science for Peace’, SfP-977998.
References
[1] M.A. McHugh, V.J. Krukonis (Eds.), Supercritical Fluid
Extraction: Principles and Practice, Butterworth-Heine-
man, Stoneham, 1993.
[2] J.L. Kendall, D.A. Canelas, J.L. Young, J.M. DeSimone,
Polymerizations in supercritical carbon dioxide, Chem.
Rev. 99 (1999) 543.
[3] P.G. Odell, G.K. Hamer, Polycarbonates via melttranses-
terification in supercritical carbon dioxide, Polym. Prep. 38
(1997) 470.
[4] Odell P.G., Polycarbonate processes with supercritical
carbon dioxide, United State Patent 5, 698, 665, 1997.
[5] E. Beckman, R.S. Porter, Crystallization of bisphenol: a
polycarbonate induced by supercritical carbon dioxide, J.
Polym. Sci. 25B (1987) 1511.
[6] A.L.C. Burke, R.D. Givence, M. Jikei, Use of CO2 in step-
growth polymerization: from plasticized polymer melts to
solid state polymerization, Polym. Prep. 38 (1997) 468.
[7] V.K. Popov, V.N. Bagratashvili, A.P. Krasnov, E.E. Said-
Galiyev, L.N. Nikitin, O.V. Afonicheva, A.D. Aliev,
Modification of tribological properties of polyarylate by
super-critical fluid impregnation of copper (II) hexafluor-
oacetyl-acetonate, Tribol. Lett. 5 (1998) 297.
[8] A.L.C. Burke, G. Maier, J.M. DeSimone, Synthesis of
polyesters in supercritical carbon dioxide, Polym. Mater.
Sci. Eng. 74 (1996) 248.
[9] S.M. Cross, D. Flowers, G. Roberts, D.J. Kiserow, J.M.
DeSimone, Solid-state polymerization of polycarbonates
using supercritical CO2, Macromolecules 32 (1999) 3170.
[10] E. Said-Galiyev, Ya. Vygodskii, L. Nikitin, R. Vinokur,
M. Gallyamov, A. Khokhlov, Synthesis of polyimides in
Erratum 129
supercritical carbon dioxide, Polym. Sci. Ser. B (Russ.) 43
(7�/8) (2000) 227.
[11] E. Said-Galiyev, Ya. Vygodskii, L. Nikitin, R. Vinokur,
M. Gallyamov, A. Khokhlov, K. Schaumburg, Synthesis
of polyimides in supercritical carbon dioxide, in: TUHH
(Ed.), Second International Meeting on High Pressure
Chemical Engineering, Hamburg, Abstracts, C4, 2001.
[12] E. Said-Galiyev, Ya. Vygodskii, L. Nikitin, R.Vinokur, M.
Gallyamov, A. Khokhlov, K. Schaumburg. Synthesis of
polyimides in supercritical carbon dioxide, in: E. Re-
verchon (Ed.), Proceedings of Sixth Conference on Super-
critical Fluids and Their Applications, 2001, p. 257.
[13] S.V. Vinogradova, V.A. Vasnev, Ya.S. Vygodskii, Cardo
polyheteroarylenes. Synthesis, properties and peculiarity,
Russ. Chem. Rev. (Uspekhi Khimii) 65 (3) (1996) 2.
[14] S.G. Kazarian, Polymer processing with supercritical
fluids, Polym. Sci. Ser. C 42 (2000) 78.
[15] T. Sarbu, T. Styranec, E. Beckman, Non fluorous poly-
mers with very high solubility in supercritical CO2 down to
low pressure, Nature 405 (2000) 165.
Erratum130