Solvent Compatibility of Poly(dimethylsiloxane)-Based Microfluidic Devices

Article abstract of DOI:10.1021/ac0346712

The compatibility of poly(dimethylsiloxane) (PDMS) with organic solvents, e.g., benzene, chloroform, acetonitrile, phenol, and methanol, was studied, which is important in considering the potential of PDMS-based microfiuidic devices in various application

Full text of DOI:10.1021/ac0346712

Anal. Chem. 2003, 75, 6544-6554
S o lve n t Co m p a t ib ilit y o f
P o ly(d im e t h yls ilo x a n e )-Ba s e d Mic ro flu id ic De vic e s
†
J e s s a m ine Ng Le e , Che olm in Pa rk, a nd Ge orge M. White s ide s *
Department of Chemistry and Chemical Biology, Harvard University, 12 Oxford Street, Cambridge, Massachusetts 02138
This paper describes the compatibility of poly(dimethyl-
siloxane) (P DMS) with organic solvents; this compatibility
is important in considering the potential of P DMS-based
microfluidic devices in a number of applications, includ-
ing that of microreactors for organic reactions. We con-
sidered three aspects of compatibility: the swelling of
P DMS in a solvent, the partitioning of solutes between a
solvent and P DMS, and the dissolution of P DMS oligo-
mers in a solvent. Of these three parameters that deter-
mine the compatibility of P DMS with a solvent, the
swelling of P DMS had the greatest influence. Experimen-
tal measurements of swelling were correlated with the
solubility parameter, δ (cal1 / 2 cm-3 / 2 ), which is based on
methane) and is not useful for manipulations requiring these
solvents. The objective of this work was to define the solvent
compatibility of PDMS as a first step in discerning what types of
solvents (other than water) can be used in microfluidic systems
fabricated in this material. It is clear that PDMS is not a universal
material and that other classes of polymers (or perhaps even glass,
despite the inconvenience of fabricating devices in rigid, brittle
materials) will be required for non- and less-polar solvents. What
are not clear are the characteristics of the solvents that are
required for a solvent to be compatible with PDMS.
The problem of solvent compatibility has three aspects: (1)
the solubility of a solvent in PDMS, since this solubility influences
the swelling of the PDMS; (2) the solubility of solutes in PDMS
(or more properly, the partition of solute between a solution and
PDMS), since loss of solute from the solvent is a concern; and
(3) the dissolution of PDMS oligomers in solvent, since these
oligomers (present as contaminants in cross-linked PDMS) are
potential contaminants in the products of reactions carried out in
PDMS.
3
the cohesive energy densities, c (cal/ cm ), of the materi-
als. Solvents that swelled P DMS the least included water,
nitromethane, dimethyl sulfoxide, ethylene glycol, per-
fluorotributylamine, perfluorodecalin, acetonitrile, and
propylene carbonate; solvents that swelled P DMS the
most were diisopropylamine, triethylamine, pentane, and
xylenes. Highly swelling solvents were useful for extracting
contaminants from bulk P DMS and for changing the
surface properties of P DMS. The feasibility of performing
organic reactions in P DMS was demonstrated by perform-
ing a Diels-Alder reaction in a microchannel.
Background on Solubility. Many parameters have been used
in calculating solubilities.7,8 We have arbitrarily chosen to use
3
cohesive energy density, c (cal/ cm ), the energy associated with
the intermolecular attractive interactions within a unit volume of
material.8 The cohesive energy density can be expressed as c
-10
)
-U/ V, where U is the molar internal energy (cal/ mol) and V
3
Several characteristics make poly(dimethylsiloxane) (PDMS)
useful in fabricating microfluidic devices intended for bioanaly-
sis: ease of fabrication (rapid prototyping, sealing, interfacing with
the user), transparency in the UV-visible regions, chemical
inertness, low polarity, low electrical conductivity, and elasticity.1
PDMS does not swell in contact with water. The cost of fabrication
in PDMS is low compared to that for many materials (e.g., glass
or silicon) commonly used in microdevices and MEMs.3
is the molar volume (cm / mol). For two materials to be soluble,
their cohesive energy densities must be similar, since this energy
must be overcome to separate the molecules of the solute to allow
the molecules of solvent to insert. For materials such as cross-
linked polymers that do not dissolve, solubility is measured by
the degree of swelling. The cohesive energy density is often
expressed in terms of the solubility parameter, or Hildebrand
value: δ ) c1/ 2 ) (-U/ V)1/ 2 (cal1/ 2 cm-3/ 2).8,11 The solubility
parameter is useful for predicting the swelling behavior of a
polymer in a solvent without knowing any other information about
the solvent.
,2
There is a growing interest in using microfluidic systems for
functions other than bioanalysis in water, including organic
synthesis in organic solvents.4 PDMS swells in contact with
nonpolar solvents (e.g., hydrocarbons, toluene, and dichloro-
-6
(
6) Tokeshi, M.; Minagawa, T.; Uchiyama, K.; Hibara, A.; Sato, K.; Hisamoto,
*
Corresponding author. Phone: 617-495-9430. Fax: 617-495-9857. E-mail:
H.; Kitamori, T. Anal. Chem. 2 0 0 2 , 74, 1565-1571.
gwhitesides@gmwgroup.harvard.edu.
Current address: Department of Metallurgical System Engineering, Yonsei
University, Seoul, 120-749, Korea.
(7) Burke, J. AIC Book Pap. Group Annu. 1 9 8 4 , 3, 13-58.
(8) Du, Y.; Xue, Y.; Frisch, H. L. Physical Properties of Polymers Handbook; AIP
Press: Woodbury, NY, 1996.
†
(
(
1) McDonald, J. C.; Whitesides, G. M. Acc. Chem. Res. 2 0 0 2 , 35, 491-499.
2) Ng, J. M. K., Gitlin, I., Stroock, A. D., and Whitesides, G. M. Electrophoresis
(9) Mark, J. E.; Eisenberg, A.; Graessley, W. W.; Mandelkern, L.; Koenig, J. L.
Physical Properties of Polymers; American Chemical Society: Washington
DC, 1984.
2
0 0 2 , 23, 3461-3473.
(
(
3) Becker, H.; Locascio, L. E. Talanta 2 0 0 2 , 56, 267-287.
4) Salimi-Moosavi, H.; Tang, T.; Harrison, D. J. J. Am. Chem. Soc. 1 9 9 7 , 119,
(10) Abboud, J.-L. M.; Notario, R. Pure Appl. Chem. 1 9 9 9 , 71, 645-718.
(11) The units for δ can be expressed as the square root of the cohesive energy
1/ 2
-3/ 2
1/ 2
8
716-8717.
density (cal cm
) or as the square root of a pressure (MPa ), where
1 cal1 cm-3/ 2 ) 0.488 88 MPa1/ 2
/ 2
.
(
5) Kakuta, M., Bessoth, F. G.; Manz, A. Chem. Rec. 2 0 0 1 , 1, 395-405.
6544 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003
10.1021/ac0346712 CCC: $25.00 © 2003 American Chemical Society
Published on Web 10/11/2003

solvent system. We therefore wished to calibrate this ranking
experimentally. Many research groups have used gravimetric
Ta b le 1 . S o lu b ilit y P a ra m e t e rs , S w e llin g Ra t io s , a n d
Dip o le Mo m e n t s o f Va rio u s S o lve n t s Us e d in Org a n ic
S yn t h e s is
methods to determine the degree of swelling of a polymer by a
solvent.1
3-17
The degree of swelling is measured by the ratio of
δa
Sb
refc
rankd
solvent
µ (D)
the mass of the swollen network and solvent combined to the
mass of the dry extracted solid. This method has two major
disadvantages. First, the mass of the combined swollen network
and solvent must be measured while in equilibrium with the
solvent in the vapor phase, so that evaporation of the solvent does
not effect the measurement. Second, the gravimetric technique
requires extra steps in its protocol compared to length measure-
ments (described below), because the unpolymerized PDMS
oligomers must first be extracted before measurement. In this
research, we measured swelling by placing a solid piece of PDMS
in a solvent for 24 h and then measuring the change in dimensions
(e.g., length) of the solid while the PDMS was still submersed in
the solvent in the liquid phase. The amount of un-cross-linked
oligomers in solid PDMS is small (∼0-5%, w/ w; see later
discussion) compared to the cross-linked network, and the
oligomers do not significantly affect the shape or length of the
cross-linked PDMS. We, therefore, did not need to extract the
perfluorotributylamine
perfluorodecalin
pentane
poly(dimethylsiloxane)
diisopropylamine
hexanes
n-heptane
triethylamine
ether
cyclohexane
trichloroethylene
dimethoxyethane (DME)
xylenes
toluene
ethyl acetate
benzene
5.6 1.00 0.0
6.6 1.00 0.0
7.1 1.44 0.0
10
10
10
32
33
3
7.3
∞
0.6-0.9 8, 14
7.3 2.13 1.2
7.3 1.35 0.0
7.4 1.34 0.0
7.5 1.58 0.7
7.5 1.38 1.1
8.2 1.33 0.0
9.2 1.34 0.9
8.8 1.32 1.6
8.9 1.41 0.3
8.9 1.31 0.4
9.0 1.18 1.8
9.2 1.28 0.0
9.2 1.39 1.0
9.3 1.21 2.8
9.3 1.38 1.7
9.5 1.03 0.9
9.5 1.22 1.7
9.9 1.22 1.6
9.9 1.06 2.9
10.0 1.16 0.5
10.6 1.06 2.2
10
10
10
8,10
10
10
10
10
10
10
8,10
10
10
10
10
8,10
10
10
8,12
10
1
8
10
2
6
11
9
12
4
13
19
14
5
chloroform
2
-butanone
18
7
tetrahydrofuran (THF)
dimethyl carbonate
chlorobenzene
methylene chloride
acetone
25
15
16
22
20
23
26
17
31
21
29
27
34
24
35
30
28
36
37
38
oligomers from the PDMS before measuring the swollen length.
The degree of swelling is expressed by the swelling ratio:9,18 S )
dioxane
pyridine
10
10
D/ D
and D
0
, where D is the length of the solid PDMS in the solvent
is the length of the dry, solid PDMS.19
N-methylpyrrolidone (NMP) 11.1 1.03 3.8
tert-butyl alcohol
acetonitrile
10.6 1.21 1.6
11.9 1.01 4.0
11.9 1.09 1.6
12.0 1.01 1.2
12.1 1.02 3.8
12.6 1.00 3.5
12.7 1.04 1.7
8,12
10
8,10
8,12
8,10
10
8,12
10
10
8,12
8,12
13,15
8,12
0
1
-propanol
RESULTS AND DISCUSSION
phenol
dimethylformamide (DMF)
nitromethane
ethyl alcohol
dimethyl sulfoxide (DMSO) 13.0 1.00 4.0
propylene carbonate
methanol
ethylene glycol
glycerol
Swelling of P DMS in Organic Solvents. Prediction of
swelling of PDMS by a solvent is important when considering
which solvents to use in performing organic syntheses in micro-
fluidic devices made in PDMS, which cosolvents to use in
separations, or which nonpolar components to expect to lose from
aqueous solution by contact with PDMS. Swelling of micro-
channels has many implications. Swelling changes the cross-
sectional area of the channel and, therefore, the rate and profile
of flow. Changes in channel dimensions due to swelling can effect
integration of the channel with components such as membranes,
detectors, mixers, or electrodes. Swelling also changes surface
properties and may cause the microfluidic device to deseal if the
PDMS is bonded to a glass substrate.
13.3 1.01 4.8
14.5 1.02 1.7
14.6 1.00 2.3
21.1 1.00 2.6
23.4 1.00 1.9
water
a
δ in units of cal cm-3/ 2. b S denotes the swelling ratio that was
1/ 2
measured experimentally; S ) D/ D0, where D is the length of PDMS
c
in the solvent and D0 is the length of the dry PDMS. References refer
d
to literature values of δ and µ. Rank refers to the order of the solvent
in decreasing swelling ability (see Figure 1).
Here, we calibrate the relation of the solubility parameter, δ,
of each solvent listed in Table 1 to the extent of swelling of PDMS.
The degree of swelling is measured by the swelling ratio, S; Table
For a binary system, the Hildebrand-Scatchard equation
relates the solubility parameters of nonpolar liquids to the enthalpy
change on mixing them: ∆H
the volume of the mixture, δ
component i, and æ
is the volume fraction of i in the mixture.8
m
) V
m
(δ
1
- δ
2 1 2 m
)2æ
æ , where V is
1
also reports the values of these ratios. Figure 1 plots the swelling
i
is the solubility parameter of the
ratio observed for each solvent against its solubility parameter.
The inset lists the solvents in descending order of solubility in
PDMS.20 As expected, solvents that have a solubility parameter
similar to that of PDMS (δ ) 7.3 cal1/ 2 cm-3/ 2) generally swell
PDMS more than solvents that have a solubility parameter
i
For two components to be soluble in one another (i.e., for swelling
to occur in a polymer-solvent system), the free energy of mixing
must be favorable, that is, ∆G
m
< 0. Since ∆G
m
) ∆H
m
- T∆S
m
,
2
2
and ∆H
where δ
m
∝ (δ
and δ
p
- δ
s
) , swelling is maximal when (δ
p
- δ
s
) is 0,
p
8
s
are the solubility parameters of the polymer and
(13) Favre, E. Eur. Polym. J. 1 9 9 6 , 32, 1183-1188.
(14) Yoo, J. S.; Kim, S. J.; Choi, J. S. J. Chem. Eng. Data 1 9 9 9 , 44, 16-22.
solvent.
Table 1 shows values of δ for a range of solvents often used in
(
15) Horkay, F.; Waldron, W. K.; McKenna, G. B. Polymer 1 9 9 5 , 36, 4525-
527.
(16) Hedden, R. C.; Wong, C.; Cohen, C. Macromolecules 1 9 9 9 , 32, 5154-5158.
4
8,10,12
organic synthesis.
Although the Hildebrand-Scatchard equa-
(
17) Hedden, R. C.; Saxena, H.; Cohen, C. Macromolecules 2 0 0 0 , 33, 8676-
684.
18) Van Krevelen, D. W. Properties of Polymers; Elsevier: New York, 1990.
tion suggests that solvents with δ similar to that of PDMS (δ )
8
1
/ 2
-3/ 2
7
.3 cal cm
) will swell PDMS effectively, the relationship
(
between δ and swelling is not linear and differs for each polymer-
(19) Swelling in terms of volume is calculated by ∆V ) S
3
.
(
20) In this paper, the number associated with the solvent, e.g., hexanes (8)
(
12) Lide, D. R. In Handbook of Chemistry and Physics, 74th ed.; Lide, D. R., Ed.;
indicates the rank of the solvent from Figure 1, in descending order of
solvability in PDMS.
CRC Press: Boca Raton, FL, 1994.
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003 6545

Fig u re 1 . Relationship between swelling ratio (S) (shown as Log (S)) of PDMS in various solvents and the solubility parameter (δ) for these
solvents. The solvents are numbered in order of decreasing swelling ability (e.g., “1” has the most swelling ability and “38” has the least swelling
ability). The dashed line indicates the solubility parameter of PDMS (δ ) 7.3 cal1 cm ). In general, a greater degree of swelling is observed
/2
-3/2
with solvents that have a value of δ similar to that of PDMS.
substantially different from that of PDMS. This relation between
S and δ, however, deviates in small but important ways. For
example, acetone and methylene chloride have indistinguishable
solubility parameters (δ ) 9.9 cal1/ 2 cm-3/ 2), but methylene
chloride swells PDMS much more than does acetone. This
observation can be explained by the polarity of the solvent. The
solubility parameter can be expressed by a sum of the dispersion
forces, polar forces, and hydrogen-bonding forces within the
methylene chloride (µ ) 1.60 D), methylene chloride has smaller
polar contributions to δ than does acetone. PDMS has a δ made
up of primarily dispersion forces; PDMS also has low polar
22
contributions (µ ) ∼0.6-0.9 D), so it is not surprising that
methylene chloride is more soluble in PDMS than is acetone.
We grouped the solvents in Figure 1 into four classifications,
depending on their solubility effect on PDMS: solvents that have
low solubility (1.00 < S < 1.10), moderate solubility (1.10 < S <
1.22), high solubility (1.28 < S < 1.58), and extreme solubility
8
2
2
material (Hansen’s total solubility parameter); that is, δ ) δ +
d
2
2
δ
p
+ δ
h
. While two solvents may have a similar total solubility
(1.58 < S < 2.13). Some solvents in which solubility parameters
were not available are listed in Tables 223 and 3
included in this analysis.
24,25
and are also
parameter, the contributions that make up this value may be
different. Two solvents with the same value of δ but a different
proportioning of these values among δ
substantial differences in swelling on exposure to PDMS. The
partitioning of δ that is most similar to that of PDMS will result
d
, δ
p
, and δ
h
may show
General Observations. Low Solubility. Low-solubility solvents
generally have δ g 9.9 cal1/ 2 cm-3/ 2. These solvents range from
water (38) to 1-propanol (21) and include most alcohols (1-
propanol, ethanol, methanol, phenol, ethylene glycol, glycerol),
nitriles (acetonitrile), disubstituted amides (NMP, DMF), and
tetrasubstituted ureas (1,1,3,3-tetramethylurea), sulfoxides (DMSO,
in the greatest swelling. Unfortunately, values of δ
are not readily available for PDMS or for many organic solvents.
In this paper, we consider the dipole moment of the solvent, µ
D) (representing the polar contributions), to help explain trends
in solubility, since µ is readily available for most materials (Table
). For example, in the case of acetone (µ ) 2.88 D) and
d p h
, δ , and δ
21
(
(
22) Diachun, N. A.; Marcus, A. H.; Hussey, D. M.; Fayer, M. D. J. Am. Chem.
Soc. 1 9 1 0 , 32; S 1 9 9 4 , 116, 1027-1032.
1
(23) Chitra, R.; Smith, P. E. J. Chem. Phys. 2 0 0 1 , 115, 5521-5530.
(
24) Simonyan, G. S.; Beileryan, N. M.; Pirumyan, E. G.; Roque, J.-P.; Boyer, B.
(
21) Literature values of δ
d
, δ
p
, and δ
h
for a limited number of polymers and
Kinet. Catal. 2 0 0 1 , 42, 474-478.
solvents are available in: Du, Y.; Xue, Y.; Frisch, H. L. Physical Properties of
Polymers Handbook; AIP Press: Woodbury, NY, 1996.
(25) Graffeuil, M.; Labarre, J. F.; Leibovici, C.; Theophanides, T. J. Chim. Phys.
Physicochim. Biol. 1 9 7 3 , 70, 1295-1298.
6546 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003

furan). These solvents tend to be nonpolar or only slightly polar,
having µ e 1 D, with the exception of ethers (µ ) 1.69 D for
tetrahydrofuran). The main contribution to δ is from dispersion
forces.
High-solubility solvents alone are generally not compatible with
PDMS microfluidic devices because of high swelling; PDMS will
absorb these solvents from a microchannel, and the polymer will
saturate with the solvent over time. A high-solubility solvent may
Ta b le 2 . S w e llin g Ra t io s o f P DMS in S o lve n t s w it h a n
Un d e fin e d S o lu b ilit y P a ra m e t e r
solvent
S
µ (D)
ref
acetic anhydride
tetramethylene sulfone
trifluoroethanol
1.02 ( 0.01
1.00 ( 0.01
1.01 ( 0.01
1.02 ( 0.01
2.80
4.68
2.46
3.47
10, 12
10
25
1,1,3,3-tetramethylurea
10
be mixed with another, low-solubility solvent to produce a mixture
that does not swell PDMS27 (or at least reduces the swelling
relative to the highly swelling solvent).28
Ta b le 3 . Eq u ilib riu m S w e llin g Ra t io s o f P DMS in
Am in e s
solvent
S
µ (D)
ref
26
High-solubility solvents are also useful for other types of
applications requiring PDMS that is free from contaminants, since
high-solubility solvents can be used to extract un-cross-linked
PDMS from the bulk polymer (see section on Dissolution of
PDMS Oligomers in Organic Solvents). For example, Graham et
al. used hexane for extracting oligomers from PDMS to avoid
transferring contaminants onto a self-assembled monolayer surface
during microcontact printing.29 Figure 1 lists other high-solubility
solvents that could be useful for this purpose.
Extreme Solubility. The extremely soluble solvents are certain
secondary and tertiary amines that swell PDMS to the greatest
extent: diisopropylamine, dipropylamine, and triethylamine. (See
section on Reactive Solvents for the swelling of PDMS in
dipropylamine.) The swelling of PDMS in diisopropylamine and
triethylamine is not surprising, as the solubility parameters of both
solvents (δ ) 7.3 and 7.5 cal1/ 2 cm-3/ 2, respectively) are similar
to that of PDMS. The extent of swelling of PDMS in amines,
however, is surprising. Table 3 lists others amines that we tested
with PDMS.
dipropylaminea
tetramethylethylenediamine
diisopropylethylamine
tributylamine
dimethyldodecylamine
ethylenediamine
∞
1.07
1.50
1.40
1.34
1.25
1.00
1.00
1.00
0.76
10
1.94
1.51
27
10
analine
pentaethylenehexamine
a
Dipropylamine swelled PDMS to S ) 2.61 after the polymer was
immersed in the solvent for 19 days. The PDMS dissolved completely
after 39 days.
tetramethylene sulfone), pyridines, and nitro compounds (nitro-
methane). These solvents generally have moderate to high polar
contributions (µ > 1.4 D). Fluorocarbons (perfluorodecalin,
trifluoroethanol, perfluorotributylamine) also do not swell PDMS,26
although they may have a small dipole moment and a solubility
parameter similar to PDMS (for example, µ ) 0 D and δ ) 6.6
1
/ 2
-3/ 2
cal cm
for perfluorodecalin). Low-solubility solvents can be
Changing the Surface P roperties of P DMS. PDMS has
used with PDMS without swelling and are, therefore, compatible
with microfluidic systems fabricated in PDMS. In principle, it
should be possible to consider carrying out synthetic reactions
in these solvents in PDMS microreactor systems.
3 2
repeating units of -OSi(CH ) - groups. This chemical structure
H2O
30
leads to a hydrophobic surface (θa ) 108°). Exposing this
surface to an air or oxygen plasma introduces silanol (Si-OH)
3
groups, destroys methyl groups (Si-CH ), and makes the surface
Moderate Solubility. We classify solvents ranging from dioxane
hydrophilic.3
0-32
PDMS that has been treated with plasma can be
(
20) to chlorobenzene (15) as moderately soluble solvents in
PDMS. These solvents have a range of solubility parameters (9.1
δ < 11.3) and dipole moments (0.46 < µ < 2.78) that overlap
with either the high- or low-solubility groups. These observations
reemphasize the fact that the proportioning of dispersion, polar,
and hydrogen forces for a particular solvent greatly influences its
ability to swell PDMS. For instance, PDMS is generally swollen
by ethers, but the swelling ratio in dioxane is moderately low (S
kept hydrophilic indefinitely by keeping the surfaces in contact
with water or polar organic solvents; otherwise, if the surface is
left in contact with air, surface rearrangements occur that bring
new hydrophobic groups to the surface to lower the surface free
energy.1,2
Extracting un-cross-linked PDMS contaminants from the bulk
polymer using organic solvents decreases the rate of regeneration
of the hydrophobic surface. Figure 2 shows a comparison of
contact angles (of water) between samples of extracted and
nonextracted PDMS that were oxidized (i.e., treated with plasma)
and not oxidized. These surfaces were left in contact with air after
initially being treated with an air plasma for 60 s.
<
)
1.16), perhaps because dioxane has a lower proportion of
dispersion forces than the high-swelling ethers. On the other hand,
PDMS is generally not soluble in alcohols, but tert-butyl alcohol
swells PDMS moderately (S ) 1.21); tert-butyl alcohol has a larger
contribution of dispersion forces to δ than do the other alcohols
tested.
High Solubility. High-solubility solvents generally have a
solubility parameter in the range 7.3-9.5 cal1/ 2 cm-3/ 2. In Figure
We used an extremely soluble solvent, diisopropylamine, to
extract the soluble components from cross-linked PDMS (1.5 cm
(27) High-solubility solvents
Generally, a mixture of solvents may or may not be soluble depending on
the proportioning of δ , δ , and δ . The problem of determining these
mixtures for PDMS is that it would require knowing δ , δ , and δ for each
are not completely incompatible with PDMS.
1
, these solvents range from benzene (14) to pentane (3) and
d
p
h
include acyclic and cyclic hydrocarbons (pentanes, hexanes,
heptane, cyclohexane), aromatic hydrocarbons (xylenes, toluene,
benzene), halogenated compounds (chloroform, trichloroethyl-
ene), and ethers (diethyl ether, dimethoxyethane, tetrahydro-
d
p
h
solvent, but these data are not available for many solvents.
(
(
28) Teas, J. P. J. Paint Technol. 1 9 6 8 , 40, 19-25.
29) Graham, D. J.; Price, D. D.; Ratner, B. D. Langmuir 2 0 0 2 , 18, 1518-1527.
(30) Morra, M.; Occhiello, E.; Marola, R.; Garbassi, F.; Humphrey, P.; Johnson,
D. J. Colloid Interface Sci. 1 9 9 0 , 137, 11-24.
(
26) Deng, T.; Ha, Y.-H.; Cheng, J. Y.; Ross, C. A.; Thomas, E. L. Langmuir 2002 ,
8, 6719-6722.
(31) Chaudhury, M. K., Whitesides, G. M. Langmuir 1 9 9 1 , 7, 1013-1025.
(32) Xia, Y.; Whitesides, G. M. Annu. Rev. Mater. Sci. 1 9 9 8 , 28, 153-184.
1
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003 6547

assembly of channels in PDMS to a flat surface of PDMS),
complicated and precise alignment of layers may take several
minutes.37 By oxidizing two pieces of extracted PDMS, we were
able to seal two pieces irreversibly as long as 90 h after plasma
treatment.38
There are several advantages to having hydrophilic surfaces
of PDMS that are slow to regenerate into the hydrophobic surface.
Hydrophilic surfaces are useful in filling and working with
microfluidic devices using water as a fluid; hydrophobic micro-
channels nucleate air bubbles easily and make it difficult to get
rid of air bubbles in the channels.39 The slow surface rearrange-
ments of extracted PDMS that has been oxidized may make it a
better substrate for surface modification (i.e., by treating the
surfaces with silanes4
0-42
or polyelectrolyte multilayers43-45) than
nonextracted PDMS.
Deswelling P DMS. PDMS that is extracted with either the
high- or extreme-solubility solvents can deswell back to its original
shape after removal of the solvent. If the PDMS is not dried evenly,
however, it tends to buckle, since uneven evaporation creates a
stress in the polymer. For instance, if the PDMS is removed from
a highly swelling solvent and left on a surface to be dried in air,
the solvent evaporates faster on the side of the polymer exposed
to air, and that side will shrink faster than the side touching the
surface, where evaporation is slow. To avoid cracking of PDMS
during deswelling, we placed the swollen polymer into decreas-
ingly soluble solvents that were miscible with the high-solubility
solvent. We deswelled PDMS by removing it from diisopropyl-
amine, placing the swollen polymer in toluene, ethyl acetate, and
then acetone (each solvent for 1 day, changing the solvent once),
drying the polymer in air, and then placing the polymer in an
oven (90 °C) for 2 days. Swelling and deswelling of PDMS did
not influence the ability of the polymer to make conformal (van
der Waals) contact with other smooth surfaces.
Fig u re 2 . Advancing contact angle measurements of water on
PDMS surfaces that were extracted or nonextracted and oxidized or
nonoxidized. Surfaces that were extracted and oxidized remained
hydrophilic in air for days; surfaces that were not extracted and
oxidized regenerated the hydrophobic surface within hours. Error bars
for the oxidized surfaces are shown and give an error of (1 standard
deviation (sample size N ) 30). Surfaces that were not oxidized (either
extracted or not extracted) remained hydrophobic. An average error
of (2° (1 standard deviation) was measured for these surfaces; error
bars were omitted on the graph for clarity.
×
1.5 cm × 0.2 cm, lwh) for 1 day at 25 °C (see section on
Dissolution of PDMS Oligomers in Organic Solvents). The
diisopropylamine was removed from the cross-linked PDMS by
immersing the samples in ethyl acetate (1 day) and acetone (2
days) and then drying in a 90 °C oven for 2 days.33 PDMS surfaces
H2O
that were extracted and then oxidized retained θa < 30° in
air for 4 days after plasma treatment. For PDMS that was oxidized
but not extracted, un-cross-linked oligomers migrated to the
surface within hours after plasma treatment. Extracted and
nonextracted PDMS that were not oxidized gave average contact
Influence of Swelling on P DMS Bonded to Glass. Glass is
sometimes used as a substrate for mounting microfluidic devices
made in PDMS.1,2,39 PDMS bonds irreversibly to glass when the
surfaces of the two are oxidized in an air plasma and then brought
together. To test the influence of swelling on the desealing of a
piece of PDMS that is chemically bonded to a glass substrate, we
placed the bonded pieces into different solvents from each solvent
category for 24 h. The low-solubility solvents (solvents 21-38 in
Figure 1) are the most compatible with PDMS and glass; these
solvents do not release (i.e., deseal) the pieces. The moderately
H2O
angles of θa ) 96° ( 2° and 98° ( 2°, respectively.
We observed that extracted and oxidized PDMS sealed better
with other similarly treated pieces of PDMS than nonextracted
PDMS that had been similarly oxidized. This property is important
for the assembly of microfluidic devices, because assembly often
involves alignment and contact of PDMS layers after the layers
have been oxidized.3
4-36
When nonextracted PDMS layers are
(
37) By placing nonextracted pieces of PDMS in a 90 °C oven for ∼10 min after
contact, we obtained irreversible sealing as long as 1 h after plasma
treatment. We found that this procedure was not always reproducible,
however.
used, irreversible sealing typically requires contact of the layers
1
within 1 min after plasma treatment. Although this amount of
time is sufficient to seal simple devices irreversibly (i.e., the
(
38) To prepare the oxidized surfaces, pieces of PDMS were oxidized in a plasma,
allowed to stand in contact with ambient air for 1 min at room temperature,
contacted, and immediately placed in a 90 °C oven for 10-20 min.
(
33) We did not test the extracted PDMS (1.5 cm × 1.5 cm × 0.2 cm, lwh) directly
for the presenece of residual diisopropylamine. We infer that diisopropyl-
amine was removed from the bulk of the PDMS, since an extraction
procedure similar to the one described here was used for the extraction of
(39) McDonald, J. C.; Duffy, D. C.; Anderson, J. R.; Chiu, D. T.; Wu, H.;
Whitesides, G. M. Electrophoresis 2 0 0 0 , 21, 27-40.
(40) Papra, A.; Bernard, A.; Juncker, D.; Larsen, N. B.; Michel, B.; Delamarche,
E. Langmuir 2 0 0 1 , 17, 4090-4095.
1
microchannels in PDMS (6.5 cm × 1.5 cm × 0.5 cm, lwh)sa H NMR
spectrum of a solution that was allowed to flow through these channels for
(41) Donzel, C.; Geissler, M.; Bernard, A.; Wolf, H.; Michel, B.; Hilborn, J.;
2
4 h showed the absense of diisopropylamine.
Delamarche, E. Adv. Mater. 2 0 0 1 , 13, 1164-1168.
(
(
34) Chiu, D. T.; Pezzoli, E.; Wu, H.; Stroock, A. D.; Whitesides, G. M. Proc.
(42) Bernard, A.; Fitzli, D.; Sonderegger, P.; Delamarche, E.; Michel, B.; Bossard,
Hans R.; Biebuyck, H. Nat. Biotechnol. 2 0 0 1 , 19, 866-869.
(43) Zhang, X.; Jiang, X. N.; Sun, C. Sens. Actuators, A 1 9 9 9 , 77, 149-156.
(44) Jiang, X.; Zheng, H.; Gourdin, S.; Hammond, P. T. Langmuir 2 0 0 2 , 18,
2607-2615.
Natl. Acad. Sci. U.S.A. 2 0 0 1 , 98, 2961-2966.
35) Anderson, J. R.; Chiu, D. T.; Jackman, R. J.; Cherniavskaya, O.; McDonald,
J. C.; Wu, H.; Whitesides, S. H.; Whitesides, G. M. Anal. Chem. 2 0 0 0 , 72,
3
158-3164.
36) Ismagilov, R. F.; Ng, J. M. K.; Kenis, P. J. A.; Whitesides, G. M. Anal. Chem.
0 0 1 , 73, 5207-5213.
(
(45) Barker, S. L. R.; Tarlov, Michael, J.; Canavan, H.; Hickman, J. J.; Locascio,
2
L. E. Anal. Chem. 2 0 0 0 , 72, 4899-4903.
6548 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003

P artitioning of Solutes between a Solvent and P DMS. The
solubility of solutes in PDMS is a factor to consider when organic
reactions are performed in systems made in PDMS, since the loss
of solute from the solvent is a concern. Whether a solute prefers
to partition into the PDMS phase or the solvent phase depends
on the relative interactions among these three components. For
this ternary system, these interactions can be described by the
Ta b le 4 . S w e llin g a n d Re a c t ivit y o f Va rio u s Ac id s ,
Ba s e s , a n d S o lve n t s Co m m o n ly Us e d in Org a n ic
S yn t h e s is
weight %a
molarityb
solvent
acetic acid
ammonium hydroxide
chloroacetic acid
dipropylamine
hydrochloric acid
nitric acid
phosphoric acid
potassium hydroxide
sodium hydroxide
sulfuric acid
S
100
29
6
100
37
70
86
28
19
5
17.0
7.5
1.0
1.00 ( 0.01
1.02 ( 0.01
1.02 ( 0.01
dissolves
1.02 ( 0.01
1.01 ( 0.01
1.01 ( 0.01
1.00 ( 0.01
1.01 ( 0.01
1.01 ( 0.01
dissolves
7.3
Flory-Huggins interaction parameters: øpolymer-solute, øpolymer-solvent
,
12.0
16.0
15.0
10.0
10.0
1.0
18.0
0.3
13.4
48
and øsolute-solvent
.
As with swelling, the Flory-Huggins interaction
parameter depends on the solubility parameter, δ, of each
2
component: ø ) (δ
1
- δ
2
) V
1
/ RT, where V
1
is the molar volume
of component 1, R is the real gas constant, and T is the
8
sulfuric acid
trifluoroacetic acid
trifluoroacetic acid
96
2
100
temperature. Maximum mixing or interaction between compo-
1.00 ( 0.01
dissolves
)2 is 0. Although
nents exists when ø is small, e.g., when (δ
1
- δ
2
these relationships exist, it is difficult to predict the partitioning
of solutes because solubility parameters for solutes are not readily
available, and it is difficult to measure each of the parameters.
We therefore wished to test the partitioning of a few solutes in
different solvents experimentally.
a
Denotes weight percent in H2O. b Molarity in units of mol/ L.
and highly soluble solvents do deseal PDMS from glass. Under
these conditions, the glass inhibits the PDMS from reaching
equilibrium swelling, since glass itself does not expand or swell
in the presence of these solvents. The PDMS is therefore under
stress; stress is relieved when the polymer deseals from glass, at
times even breaking the glass or tearing the PDMS.
We tested the partitioning of a few UV-active compounds
rhodamine B chloride, fluorescein, and dansyl chloride) in
different mixtures of PDMS prepolymer and solvent.
(
Swelling of P DMS in Acids and Bases. Swelling of a
polymer in an acid or a base cannot be predicted by considering
solubility parameters, as the solubility parameter does not account
for ionic interactions and chemical reactions. We therefore tested
the swelling of PDMS in common acids and bases used in organic
synthesis. Table 4 lists these acids and bases, as well as the
measured swelling ratios of PDMS in these solutions after 24 h.
PDMS did not swell considerably in any of the acids and bases in
the concentrations listed; PDMS is therefore compatible with these
solutions (except for the ones in which the PDMS dissolved, see
below) for organic synthesis.
Reactive Solvents. We define reactive solvents as ones that
react with PDMS to break the polymer down into smaller subunits
(
Table 4). The first group of solvents that reacted with PDMS
are certain acids: concentrated sulfuric acid and trifluoroacetic
acid. Both of these acids dissolved PDMS and left a white
precipitate after being immersed for 24 h. The products of the
reactions consisted of oligomers with different numbers of
To form a mixture, and not a solution or a suspension with
the PDMS prepolymer, the compounds to be tested were dissolved
in a low-solubility solvent (e.g., a solvent that is immiscible with
PDMS prepolymer). The ideal choice of low-solubility solvent
3 2
dimethylsiloxane subunits, -(CH ) SiO-, determined by mass
spectrometry and IR.
(
from Figure 1) had three characteristics: (1) the solvent solvated
each of the compounds; (2) the solvent did not react with the
solute, nor quench the fluorescence of the solute in the presence
of PDMS prepolymer; and (3) the solvent did not cause the PDMS
prepolymer to cross-link. Solvents that best fit these criteria were
water, propylene carbonate, nitromethane, and acetonitrile.49
Partitioning experiments were performed with solutions of
rhodamine, fluorescein, and dansyl chloride using the concentra-
tion of solute in each solvent that gave an absorbance (A) value
of ∼0.8. Equal amounts of solvent and PDMS prepolymer were
mixed with vigorous shaking for 2 min and then allowed to
Certain organic reagents react with PDMS. A 1 M tetrabutyl-
ammonium fluoride (TBAF) solution in THF dissolves PDMS,
most likely by breaking Si-O bonds and forming Si-F bonds, in
a reaction similar to the etching of glass that occurs in hydrogen
fluoride.46 Dipropylamine also dissolves PDMS, although more
slowly than TBAF. Dipropylamine first swells PDMS over a period
of weeks before dissolving47 the PDMS into dimethylsiloxane
3 2
subunits, -(CH ) SiO- (determined by mass spectrometry); we
have not determined the processes involved in this solublization.
Reactive solvents can be used as etching agents for PDMS.
(
(
46) Arias, F.; Oliver, S. R. J.; Xu, B.; Holmlin, R. E.; Whitesides, G. M. J.
Micromechan. Syst. 2 0 0 1 , 10, 107-112.
47) Pieces of PDMS began losing their shape and became gellike after being
(48) Young, R. J. Introduction to Polymers; Chapman and Hall: New York, 1991.
(49) From the low-solubility solvents listed in Figure 1, pyridine, NMP, DMF,
and DMSO cross-linked the PDMS prepolymer into a clumpy gel upon
mixing.
immersed in dipropylamine for 36 days.
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003 6549

Ta b le 5 . P a rt it io n in g o f Org a n ic Co m p o u n d s in S o lve n t /P DMS Mix t u re s
partition ratios (Asolvent/ Asolvent + APDMS) (for given solvent)
compound
water
propylene carbonate
nitromethane
ethanol
acetonitrile
rhodamine
fluorescein
dansyl chloride
0.4 ( 0.1
1.00 ( 0.04
a
0.83 ( 0.07
a
1.00 ( 0.01
0.83 ( 0.08
a
1.00 ( 0.01
0.97 ( 0.03c
0.71 ( 0.02
b
b
a
c
1.00 ( 0.01
a
Compound did not dissolve in solvent. b Fluorescence of compound was quenched in the solvent/ PDMS mixture. c Only the solvent phase
was tested, since the PDMS phase was turbid after mixing with solvent.
partition in the dark for 24 h. The absorbencies of the solute in
the solvent and PDMS phases were measured; here, we present
the degree of partitioning of the solute in each of the phases as
a ratio of absorbencies, Asolvent/ (Asolvent + APDMS) e 1 (Table 5).
In these experiments, øsolute-solvent is small, since we chose
solvents in which the solute dissolved and made a clear solution.
We thus assume that øsolute-solvent is similar for the same compound
in the different solvents we tested. For the same compound,
solutions of the compound in different solvents have the same
ø
polymer-solute, since the polymer and solute components do not
change. We therefore compare the degree of partitioning with,
polymer-solvent, which we estimate by the difference in δ between
the polymer and solvent. For the partitioning of rhodamine (Table
), there is no obvious trend between solubility parameter and
ø
5
partition ratio. (Note that the solvents in Table 5 are listed from
left to right in order of decreasing solubility parameter.) The
partitioning of dansyl chloride did not change with slight variations
in solubility parameter of the solvents tested.
Fig u re 3 . Relationship between the percent extracted PDMS (w/
w) and the swelling ratio (S) of various solvents in Table 1.
Although we did not see a trend between solubility parameter
and partition ratio for the solvents we tested, we generally do not
expect the partitioning of solutes in PDMS to be a deterring factor
for performing organic reactions in microfluidic channels made
network. The cross-linked (cured) PDMS, therefore, contains un-
cross-linked, low molecular weight oligomers that are included
in the bulk polymer. When PDMS is in contact with an organic
solvent, that solvent may extract these un-cross-linked oligomers
from the bulk. We wished to determine the extent of dissolution
of PDMS in a solvent, because these oligomers might contaminate
the products of reactions carried out in PDMS.
in PDMS, since the range of compatibilities for polymer-solute
systems is much smaller than that of solvent-solute systems.50
For instance, the critical parameter for the complete mixing of
two small molecules (solvent and solute) in the regular solution
model is R ) 2, such that if R < 2, the two molecules will mix
and make a clear solution; otherwise, they will partition.51 The
We tested the extent of dissolution of PDMS in various solvents
by immersing pieces of PDMS (5 cm × 3.5 cm × 0.3 cm, lwh) in
the solvents for 24 h and determining the amount of un-cross-
linked PDMS by measuring the difference in the weight of the
polymer before and after extraction (see Experimental Section).
The extent of dissolution of PDMS was also calculated by
critical parameter for a polymer-solute mixture in the Flory-
Huggins model, however, is ø ) 0.5; thus, for a polymer to mix
homogeneously with a solute requires ø < 0.5, a smaller range
5
2
than for a solvent-solute mixture. Because the range of values
that is required to make a mixture homogeneous is smaller for a
polymer-solute system than for a solvent-solute system, it is
generally more common for a solute to be miscible with a solvent
1
analyzing the solvent used for extracting the PDMS by H NMR;
this method gave values in agreement with the values obtained
by weight measurements within experimental error. Figure 3 plots
the weight percent of PDMS extracted against the swelling ratio
than a polymer.8
,48
Extraction of P DMS Oligomers into Organic Solvents.
When PDMS oligomers are cross-linked to form the polymer, not
all of the oligomer strands are incorporated into the cross-linked
(
52) Both R and ø are temperature-dependent, dimensionless parameters that
represent the enthalpy term in the Gibbs free energy equation. The
difference between the parameters is that R represents the pair-pair
interaction energy in a solvent-solute solution (i.e., the solvent molecule-
solute molecule interaction), while ø represents two types of interactions in
a polymer-solvent solution: pair-pair interaction energy (i.e., the segment-
segment interaction of the polymer) and the polymer volume and solvent
volume interaction. Since ø involves volume interactions, it contains an
entropic contribution, making it not as straightforward to interpret as the R
term. For general discussion in this paper, however, we compare the values
of R and ø directly. (For further discussion, refer to the section on Flory-
Huggins Theory in: Young, R. J. Introduction to Polymers; Chapman and
Hall: New York, 1991.)
(
50) There are some solutes, however, that are more likely than others to partition
into PDMS. For example, solutes that have a chemical structure similar to
the extremely soluble solvents (e.g., diisopropylamine and triethylamine)
will likely partition into the polymer more than solutes that have charac-
teristics of the low-solubility solvents (such as alcohols and ketones). The
high solubility of amines in PDMS may explain why the partition ratio of
rhodamine was quite low in the PDMS-water mixture (i.e., rhodamine
favored the PDMS phase over the water phase, Table 5).
(
51) Gaskell, D. R. Introduction to the thermodynamics of materials; Taylor &
Francis: Washington, DC, 1995.
6550 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003

Fig u re 4 . Schematic diagram of the microfluidic device used for the Diels-Alder reaction. The channels contain a chaotic advective mixer
CAM)sridges that change their orientation with respect to the center of the channel every half-cycle. The streamlines of flow in the channel
(
cross section are also shown. We inject solutions of the reagents 4-phenyl-1,2,4-triazoline-3,5-dione (position 1) and ergosterol (position 2) into
the inlets using a syringe pump that is connected to the device via polyethylene tubing. The middle solvent inlet contains acetone and is only
used to fill and rinse the channels. The adduct collects into a vial at the outlet (position 3) via polyethylene tubing.
of PDMS in each solvent. Generally, the amount of extracted
PDMS increases as the swelling ratio increases. We anticipated
this trend.53
extracted was negligible (<10-3% (w/ w) of the total amount of
solvent allowed to flow through the channel). We therefore do
not believe that contamination of the product with PDMS will be
an issue when performing organic reactions in microchannels
fabricated of PDMS that have previously been extracted with a
highly soluble solvent.
We suspect that the extent of dissolution of PDMS in a
microchannel will be less than the amounts reported in Figure 3.
For example, if the low-solubility solvents 21-38 are used in
microfluidic applications, the extent of dissolution of PDMS in a
microchannel will be <2.5% (w/ w) (i.e., less than the highest
swelling solvent of the low-solubility group, 1-propanol). In a
microchannel, the surface area in contact with the solvent is
smaller (on the order of 10- smaller for a 25-cm-long channel
with a width and height of 100 µm) than the surface areas used
in the extractions in Figure 3. A smaller surface area in contact
with a low-solubility solvent makes extraction slower.
Applications in Synthesis. We demonstrate the compatibility
of PDMS with organic solvents by performing a Diels-Alder
5
4,55
reaction in a microfluidic channel made in PDMS (Figure 4).
The reagents 4-phenyl-1,2,4-triazoline-3,5-dione (14.9 mg, 0.085
mmol)
1
We compared the dissolution of a PDMS slab extracted with
∼
400 mL of 1-propanol (2.5% extracted PDMS (w/ w) with 24 h
of mixing, Figure 3) with the dissolution of PDMS from 1-propanol
flowing through a microfluidic channel (dimensions 25 cm × 75
µm × 75 µm, lwh) that was bonded to glass. After allowing a
solution of 1-propanol to flow through the channel at 5 µL/ min
for 24 h, the amount of un-cross-linked PDMS that was collected
in the solution at the outlet was only 0.02% of the original weight
1
of the PDMS, determined by H NMR. This amount of un-cross-
linked PDMS correlates to 0.05% (w/ w) of the total amount of
solvent allowed to flow through the channel.
To reduce or eliminate contamination from PDMS oligomers,
the microfluidic channels that are embedded in PDMS can first
be extracted with a highly soluble solvent (e.g., diisopropylamine
or pentane) and then deswelled back to the original shape of the
polymer. We performed this extraction procedure (using diiso-
propylamine) on a microfluidic network made in PDMS before
testing the dissolution of PDMS. A solution of 1-propanol was then
allowed to flow under the same conditions as those used with
the device made in unextracted PDMS. The amount of un-cross-
linked PDMS collected at the outlet of the device that was
in acetone (0.2 mL) and ergosterol (22.3 mg, 0.056 mmol) in
1:1 acetone/ benzene (0.9 mL) were injected into the inlets
(positions 1 and 2) with flow rates of 11 and 44 µL/ min,
respectively. The reaction took place along the channel, with
mixing aided by a chaotic advective mixer (CAM)56 that had been
incorporated into the channel. The solvent inlet contained acetone
(
53) The amount of un-cross-linked PDMS in the bulk polymer depends on the
ratio of PDMS prepolymer to cross-linking agent (a 10:1 (w/ w) ratio is
recommended), as well as the degree of mixing of the two components. A
well-mixed solution of PDMS will contain less un-cross-linked PDMS than
a solution that is not well-mixed. The amount of extracted PDMS depends
on the efficiency of the solvent to swell the polymer.
(54) Gilani, S. S. H.; Triggle, D. J. J. Org. Chem. 1 9 6 6 , 31, 2397-2398.
(55) Morris, D. S.; Williams, D. H.; Norris, A. F. J. Org. Chem. 1 9 8 1 , 46, 3422-
3428.
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003 6551

and was only opened initially to fill the channels and to rinse the
channels when the injections of reagents were stopped; this inlet
remained closed during the time when the reagents were injected.
The dione solution turned from bright red to a yellow-pink upon
reaction with ergosterol in the channel; this color change was
CONCLUSIONS
We have determined a range of solvents that is compatible
with a range of operations in PDMS, including performing organic
synthesis in microfluidic channels made in PDMS. Of the three
parameters that determine the compatibility of PDMS with a
solventsi.e., the swelling of PDMS in the solvent, the partition of
solute between a solvent and PDMS, and the dissolution of PDMS
oligomers in the solventsthe swelling of PDMS had the greatest
influence. Generally, low-solubility (weakly swelling) solvents are
most compatible with performing organic reactions in microfluidic
channels made in PDMS, although some mixtures of low- and
high-solubility solvents are also compatible. We demonstrated this
compatibility by performing a Diels-Alder reaction in PDMS
microchannels. Acids and bases do not swell PDMS, but reactions
that require amines or certain strong acids are not compatible,
as these reagents dissolved the polymer.
observed to take place within seconds inside the channel and not
at the outlet.57 The reaction mixture was collected at the outlet
(
position 3) into a vial via polyethylene tubing. The adduct was
identified in the 1H NMR spectrum of the reaction mixture by
the appearance of two peaks from the hydrogen of cyclohexene
(
1 H, d, 6.2 ppm; 1 H, d, 6.4 ppm); the formation of these peaks
is typical of Diels-Alder reactions.54 The two peaks from cyclo-
hexadiene of ergosterol (1H, d, 5.4 ppm; 1H, d, 5.6 ppm) did not
appear in the spectrum, indicating a 100% conversion of the
limiting reagent (ergosterol) to the product. Unreacted 4-phenyl-
1
,2,4-triazoline-3,5-dione was also identified in the reaction mixture
1
High-solubility solvents are useful for extracting PDMS oligo-
mers from the bulk of the cross-linked polymer. This extraction
process helps to reduce or eliminate contamination (from PDMS
oligomers) of solutions that flow through microchannels made in
PDMS. Since PDMS deswells approximately back into its original
shape when the swollen polymer is cycled through decreasingly
soluble solvents, the shape of the microchannels is not disturbed
after extraction. We also found that extraction of PDMS changes
the surface properties of PDMS: surfaces of extracted PDMS that
by H NMR.
In this experiment, we used a microchannel that was first
extracted with pentane to remove the un-cross-linked PDMS
1
oligomers. H NMR spectra of the reaction mixture collected at
the outlet confirmed that the presence of PDMS was negligible.
We also did not encounter problems of swelling of PDMS, even
though we used benzene, a highly swelling solvent.58 The
microfluidic channel did not deseal from the glass substrate during
the time that was allowed for the reaction to occur (15 min).
Although we did not integrate analyte detection into this
microfluidic device, detection methods such as mass spectrom-
etry,5,59,60 or ones based on fluorescence,61,62 electrochemistry,63-65
or absorbance,66,67 can be coupled to microfluidic systems for
identifying reagents and products of organic reactions. Other steps
that may be important for chemical synthesis, such as heating,
cooling, and filtration, may also be integrated into microfluidic
systems made in PDMS, since PDMS is a material that allows
easy integration and interfacing of components.1,2
H2O
were oxidized with an air plasma retained θa < 30° in air for 4
days after plasma treatment and sealed better than nonextracted
PDMS. Extracted PDMS may be useful for those wanting
hydrophilic channels for microfluidics or those interested in
modifying the surface of PDMS.
Although problems of swelling and compatibility of organic
solvents is less of a concern in microfluidic systems fabricated in
glass or silicon, the cost of fabrication in these materials is high.
The cost and ease of fabrication in PDMS is low, but PDMS is
limited to certain ranges of solvents. As interest in performing
organic reactions on-chip continues to grow, it seems likely that
PDMS will be used in fabricating microreactors for organic
reactions that involve solvents compatible with PDMS. Reactions
requiring highly and extremely soluble solvents may require glass
or silicon.
(
(
56) Stroock, A. D.; Dertinger, S. K. W.; Ajdari, A.; Mezic, I.; Stone, H. A.;
Whitesides, G. M. Science 2 0 0 2 , 295, 647-651.
57) Since the reagents flow at a rate of ∼44 mL/ min and are adequately mixed
in the channel, it should take ∼40 s for the reagents to reach the outlet for
a channel with dimensions 200 µm × 75 µm × 20 cm (whl). The time for
reaction is , 40 s (Gilani, S. S. H.; Triggle, D. J. J. Org. Chem. 1 9 6 6 , 31,
EXPERIMENTAL SECTION
2
397-2398). Thus, the reaction should take place in the channel and not at
the outlet. Observations of color change within the channel, and not at the
outlet, agree with this analysis.
58) We may not have observed swelling of PDMS with benzene in this
experiment because of the short length of time in which the solvent was in
contact with the polymer (15 min for the experiment performed here, versus
Materials. Sylgard 184 Silicone, a two-part PDMS elastomer,
was purchased from Essex Brownell (Edison, NJ). For all
experiments requiring solid PDMS, we used a 10:1 (by weight)
mixture of PDMS base/ curing agent that was degassed under
vacuum and cured at 70 °C for 24 h. Solvents were obtained from
Sigma-Aldrich Co. (St. Louis, MO), Fisher Scientific Co. (Pitts-
burgh, PA), and Mallinckrodt, Inc. (Chicago, IL) and used as
received.
Swelling Measurements. We measured swelling (at 25 °C)
by comparing the lengths of solid pieces of PDMS before and
after being immersed in a solvent. The PDMS pieces were made
by soft lithography and are in the shape of hexagons;32,68 the length
from one edge of the hexagon to the opposite edge was measured
(
2
4 h for the swelling experiment performed in Figure 1, where swelling
was observed).
(
(
(
(
(
59) Kameoka, J.; Craighead, H. G.; Zhang, H.; Henion, J. Anal. Chem 2 0 0 1 ,
7
3, 1935-1941.
60) Roulet, J.-C. V.; R., Herzig, H. P.; Verpoorte, E.; de Rooij, N. F.; Daendliker,
R. Anal. Chem 2 0 0 2 , 74, 3400-3407.
61) Chabinyc, M. L.; Chiu, D. T.; McDonald, J. C.; Stroock, A. D.; Christian, J.
F.; Karger, A. M.; Whitesides, G. M. Anal. Chem. 2 0 0 1 , 73, 4491-4498.
62) Liang, Z.; Chiem, N.; Ocvirk, G.; Tang, T.; Fluri, K.; Harrison, D. J. Anal.
Chem. 1 9 9 6 , 68, 1040-1046.
63) Zhan, W.; Alvarez, J.; Sun, L.; Crooks, R. M. Anal. Chem 2 0 0 3 , 75, 1233-
1
238.
(
(
(
64) Hebert, N.; Kuhr, W.; Brazill, S. A. Electrophoresis 2 0 0 2 , 23, 3750-3759.
65) Wang, J. Talanta 2 0 0 2 , 56, 223-231.
66) Adams, M. L.; Enzelberger, M.; Quake, S.; Scherer, A. Sens. Actuators, A
(
typically, 4 mm from one edge to the opposite edge, and 1 mm
thick for unswollen pieces). We immersed these PDMS pieces in
2
0 0 3 , 104, 25-31.
(
67) Bowden, M. a. D., D. Sens. Actuators, B 2 0 0 3 , B90, 170-174.
(68) Bowden, N.; Oliver, S. R. J.; Whitesides, G. M. J. Phys. Chem. B 2 0 0 0 , 104.
6552 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003

each solvent for 24 h at 25 °C; under these conditions, swelling
reached equilibrium (i.e., the dimensions of the PDMS solid did
not change with time).69 The pieces were imaged while immersed
in the solvent, using a CCD camera connected to a stereoscope;
the lengths of the pieces were measured directly on the digitized
image, using Nikon ACT-1 Version 2.12 software (Nikon Corp.).
of 1-propanol in the channel (25 cm × 200 µm × 200 µm, lwh).
Microfluidic channels were made by soft lithographic procedures
described elsewhere.32,39 The solution was collected at the end of
the channel, and the solvent was evaporated under vacuum. The
amount of un-cross-linked PDMS was determined by adding a
1
known amount of internal standard and taking a H NMR in CDCl .
3
(
If the PDMS piece was removed from the solvent, we found that
The amount of un-cross-linked PDMS was 0.02% (w/ w) of the
original weight of PDMS, or 0.05% (w/ w) of the total amount of
solvent flowed through the channel.
the dimensions of the piece decreased within seconds as the
solvent evaporated from the PDMS.) We measured the lengths
of five different pieces of PDMS for each solvent; an average of
these measurements gave the reported value with a typical relative
error of (1% for the different solvents.
Reactive Solvents. To determine which acids and bases
reacted with PDMS, pieces of PDMS in the shapes of hexagons
(
4 mm in length from one edge to the opposite edge) were
Contact Angle Measurements. Advancing contact angles
were measured on static drops of water using a Ram e´ -Hart
goniometer with a Matrix Technologies Electrapipet to control
the advancement of the drop. Typically, N ) 30 measurements
were taken for each of the data points in Figure 2. We measured
the advancing contact angles of water on pieces of PDMS (1.5
cm × 1.5 cm × 0.2 cm, lwh) that were extracted or nonextracted
and oxidized or nonoxidized. The pieces of extracted PDMS were
extracted (at 25 °C) for 1 day in ∼100 mL of diisopropylamine
with stirring, deswelled in stirred solutions of ∼100 mL of ethyl
acetate for 1 day and then acetone for 2 days (∼100 mL each
day; solvent was changed after 1 day), and dried in an oven (90
immersed in 10 mL of acid or base for 24 h at 25 °C. The solvents
that reacted with PDMS (18.0 M sulfuric acid and 13.4 M
trifluoroacetic acid) formed a white precipitate after this time. To
determine the product of the reaction of acid with PDMS, the
precipitate was rinsed thoroughly with water and then dried in
an oven (70 °C) for 24 h. Mass spectra and IR spectra of the
precipitate give data that are consistent with the hypothesis that
the products of both reactions consisted of low molecular weight
3 3 3 2 x 3 3
oligomers having the structure (CH ) Si[OSi(CH ) ] OSi(CH ) .
P artitioning. For partitioning experiments, we used one part
(
the PDMS base) of a two-part PDMS elastomer, Sylgard 184
Silicone, from Essex Brownell. Rhodamine B chloride and
fluorescein were purchased from Sigma-Aldrich Co., dansyl
chloride was purchased from Fluka, and all solvents were
purchased from Sigma-Aldrich Co. and used as received.
Partitioning of UV-active compounds (rhodamine, fluorescein,
dansyl chloride) was tested in mixtures containing PDMS pre-
polymer and solvent. Solutions of each compound were made in
water, propylene carbonate, nitromethane, ethanol, and aceto-
nitrile. The wavelengths that gave maximum absorption, λabs, for
each compound varied slightly in each solvent and were deter-
mined to be ∼554-588 nm for rhodamine, 490 nm for fluorescein,
and 373 nm for dansyl chloride. Calibration curves containing a
range of concentrations of each solute that gave absorbance (A)
measurements (at λabs) between 0 and ∼1.5 were determined
using a HP 8453 UV-visible spectrophotometer (Hewlett-Packard)
and UV-visible ChemStation software (Agilent Technologies).
We chose to perform partitioning experiments with the
concentration of solute that gave A ∼0.8 for each solvent. These
concentrations were on the order of 0.01 mM (depending on the
solvent) for rhodamine, 0.01 mM for fluorescein, and 0.2 mM for
dansyl chloride. To a vial was added 3.0 g of the solvent containing
the solute and 3.0 g of PDMS prepolymer. The contents were
mixed for 2 min using a Vortex-Genie 2 (VWR Scientific) set to
maximum. The mixture was covered from exposure to light, and
the phases were allowed to separate for 24 h. The PDMS phase
was removed from the solvent phase using a pipet and then placed
in a cuvette for UV-visible; this phase was degassed under
vacuum for 24 h (in the dark) to remove the air bubbles before
performing absorbance measurements. Measurements of ab-
sorbance were obtained for both the solvent phase and PDMS
phase. Three sets of partitioning experiments were performed for
each of the solute-solvent combinations.
°
C) for at least 2 days to remove the residual solvent from the
bulk of the polymer. The oxidized surfaces were oxidized for 60
s in a SPI Plasma Prep II plasma cleaner that was used under
vacuum (∼2 mTorr) (SPI Supplies, West Chester, PA).
Swelling of P DMS Bonded to Glass. To test whether a
solvent desealed a piece of PDMS that was chemically bonded to
a glass substrate, pieces of PDMS (4 cm × 3 cm × 0.3 cm, lwh)
and glass slides (7.5 cm × 2.5 cm × 0.1 cm, lwh) were oxidized
in an air plasma. The surfaces that were exposed to the plasma
were brought together and sealed irreversibly by applying a small
pressure for 1 min. These bonded pieces were then placed into
the solvent for 24 h without stirring at 25 °C.
Extraction of Un-Cross-Linked P DMS. To test the extent
of dissolution of PDMS in a solvent (Figure 3), we immersed
pieces of PDMS (5 cm × 3.5 cm × 0.3 cm, lwh) into ∼400 mL of
the solvent for 24 h at 25 °C while the solvent was stirred. The
percent extraction (w/ w) was determined by dividing the differ-
ence between the weight of the PDMS before and after extraction
by the original weight of the PDMS. We also analyzed the solvent
used for extracting the PDMS by evaporating off the solvent under
vacuum, adding a known amount of internal standard (DMSO),
1
.70 (The
and taking a H NMR of the extracted PDMS in CDCl
3
PDMS peak had a chemical shift of σ ) 0.1 ppm.) The amount of
extracted PDMS calculated by this method agreed with the values
obtained by weight measurements, within experimental error.
To test the extent of extraction of PDMS by a solvent flowing
through a microfluidic channel, we used a Harvard Apparatus 2000
syringe pump to produce a steady flow rate (5 µL/ min for 24 h)
(
69) Typically, equilibrium swelling was reached within 1 h of being immersed
in the solvent. For the amines, however, equilibrium swelling time varied
from days to weeks.
(
70) We assumed the PDMS to have an average molecular weight of 50 000 based
on the viscosity of the polymer immediately after mixing with the curing
agent (4000 mPa/ s) (www.dowcorning.com) and the relationships between
viscosity and molecular weight for PDMS (https:/ / www.sigma-aldrich.com/
aldrich/ brochure/ al•pp•viscosity.pdf).
Organic Synthesis in Microfluidic Channels. All reagents
were purchased from Sigma-Aldrich Co. and used without further
purification. Microfluidic channels with CAM surface features were
Analytical Chemistry, Vol. 75, No. 23, December 1, 2003 6553

fabricated in PDMS according to reported procedures.56 The
PDMS channels were extracted with pentane for 2 days, deswelled
in toluene for 1 day, ethyl acetate for 1 day, and acetone for 1
day, and then dried in an oven (90 °C) for 2 days. The PDMS
was oxidized in an air plasma for 60 s and then bonded irreversibly
to glass.
6.4 ppm). We estimate 100% conversion of the limiting reagent
(ergosterol) to the product based on the disappearance of the two
peaks from cyclohexadiene in ergosterol (1H, d, 5.4 ppm; 1H, d,
5.6 ppm). An excess of ∼47% of 4-phenyl-1,2,4-triazoline-3,5-dione
in the end reaction mixture was calculated by comparing the ratio
of integration of the peaks from the phenyl group on 4-phenyl-
1,2,4-triazoline-3,5-dione (∼7.3 H (calculated), m, 7.4 ppm) to the
expected integration (5 H, m, 7.4 ppm). This value is in agreement
with the 52% excess of starting reagent of 4-phenyl-1,2,4-triazoline-
3,5-dione, compared to the initial amount of ergosterol.
We filled 0.250- and 1-mL syringes with solutions of 4-phenyl-
1
,2,4-triazoline-3,5-dione (14.9 mg, 0.085 mmol) in acetone (0.2
mL) and ergosterol (22.3 mg, 0.056 mmol) in a 1:1 solution of
acetone/ benzene (0.9 mL). The solutions were injected into the
channels via polyethylene tubing, with flow rates of 0.011 (for the
ACKNOWLEDGMENT
0
1
.25-mL syringe) and 0.044 mL/ min (for the 1-mL syringe) for
5 min using a Harvard Apparatus 2000 syringe pump. The
Financial support was provided by DARPA/ NSF (ECS-
0004030) and NIH (GM 65364). J.N.L. acknowledges Elisa Franqui
reaction proceeded inside the microchannels by evidence of a
color change from the bright red 4-phenyl-1,2,4-triazoline-3,5-dione
to a yellow-pink (product). The reaction mixture was collected at
for her assistance with swelling experiments.
Received for review June 20, 2003. Accepted September
2, 2003.
the outlet via polyethylene tubing into a vial; the adduct was
1
identified by H NMR in CDCl
3
by the appearance of two peaks
due to the hydrogen from cyclohexene (1 H, d, 6.2 ppm; 1 H, d,
AC0346712
6554 Analytical Chemistry, Vol. 75, No. 23, December 1, 2003