ARTICLE

pubs.acs.org/JACS

Flow-Directed Block Copolymer Micelle Morphologies via Microfluidic

Self-Assembly
Chih-Wei Wang,† David Sinton,‡ and Matthew G. Moffitt*,†
†
Department of Chemistry, University of Victoria, P.O. Box 3065, Victoria, BC, Canada V8W 3V6
‡
Department of Mechanical and Industrial Engineering, University of Toronto, 5 King’s College Road, Toronto, ON, Canada, M5S 3G8
bS Supporting Information
ABSTRACT: The self-assembly of amphiphilic block copoly-
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.

mers in a gas
liquid microfluidic reactor produces variable,

flow-directed micellar morphologies entirely different from off-
chip equilibrium structures. A polystyrene-block-poly(acrylic
Downloaded via UNIV DE BUENOS AIRES on May 16, 2019 at 14:22:23 (UTC).

acid) copolymer, which forms exclusively spheres off-chip, gen-

erates kinetic cylinders, Y-junctions, bilayers, and networks by a
mechanism of collision
coalescence enabled by strong and
localized on-chip shear fields. Variation in the size and relative amount of flow-directed nanostructures is achieved by changing the
water content and flow rate. These results demonstrate on-chip processing routes to specific functional colloidal nanostructures.

’ INTRODUCTION patterned substrates,27
32 there are very few examples of top-

The solution self-assembly of amphiphilic block copolymers down morphological control in block copolymer colloids.33,34
into colloidal nanostructures, termed block copolymer micelles, In one example, Wu and co-workers extruded a block copolymer
has spurred extensive fundamental and applied research in recent colloid through a nanoporous template to induce a sphere-
years.1
20 Of particular interest is the plethora of diverse micelle to-cylinder micelle transition, with the resulting cylinders revert-
morphologies, including spheres, cylinders, and vesicles, that can ing back to spheres upon standing for several hours.33 More
spontaneously form in such systems, as a result of the exquisite recently, work from the Winnik group reported that block
and complex interplay of thermodynamic forces. A central focus copolymer spherical micelles decorated with CdSe quantum
of work in this area has been the understanding and control of dots could be converted to a wormlike network by rapid mag-
these varied morphologies, because the size and shape of micellar netic stirring, whereas similar stirring conditions yielded only
structures will profoundly influence their effectiveness for spe- spheres in the absence of quantum dots.34
cific applications, including drug delivery,21,22 sensing,23 and Microfluidic reactors have been studied as environments for
medical imaging.24
26 Under equilibrium conditions, the mor- the self-assembly of polymers35
43 and other molecular building
phologies of block copolymer micelles are governed by a fine blocks,44
46 because the associated nanoscale volumes and
balance of interfacial tension, chain stretching within the micelle laminar flow enable fine-tuning of local concentrations of re-
core, and repulsion among solubilized coronal chains.1,2,5,7,8,10 agents, mixing rates, and shear forces. We recently applied a
However, along with such thermodynamic factors, kinetic effects gas
liquid multiphase microfluidic reactor, in which compart-
arising from a combination of slow chain dynamics and high mentalized liquid plugs are segmented by a regular stream of gas
interfacial tension can also contribute significantly to the bubbles, in order to form spherical colloidal aggregates of
products of self-assembly, often leading to structurally com- polymer-coated quantum dots.41,42 In that work, we demon-
plex and kinetically stable morphologies that are not globally strated that flow-variable high shear regions in the corners of the
equilibrated.5,10,13,14 liquid plugs strongly influenced particle sizes via competing
Conventional experimental control of block copolymer mor- mechanisms of coalescence and particle break-up.42 In the
phologies involves variation in chemical parameters that are present paper, we demonstrate that the solution self-assembly
known to directly influence intermolecular interactions (e.g., of amphiphilic block copolymers in a similar multiphase reactor
block copolymer composition,1,2 choice of solvent,8 polymer leads to flow-variable micelle morphologies that are different
concentration,11 pH, and ionic strength4). In contrast to this from equilibrium structures produced by off-chip self-assembly
bottom-up approach, the top-down control of block copolymer under identical chemical conditions. The resulting shear-induced
morphologies through templates or external force fields (e.g., nanostructures are kinetically controlled by the on-chip flow
shear and magnetic and electric fields) opens the door to environment but are stabilized by kinetic trapping after transfer
engineering micellar structures under a given set of chemical con- from the reactor into aqueous environments. This approach
ditions by “dialing in” specific processing parameters. Although
such directed assembly has been widely applied in block co- Received: July 19, 2011
polymer thin films using chemically and topologically defined Published: October 12, 2011

r 2011 American Chemical Society 18853 dx.doi.org/10.1021/ja2067252 | J. Am. Chem. Soc. 2011, 133, 18853–18864
 Edited by Foxit Reader
Copyright(C) by Foxit Software Company,2005-2008
Journal of the American Chemical Society For Evaluation Only.
ARTICLE

provides a convenient top-down handle on block copolymer to remove all moisture. SU-8 films, 150 μm thick, were spin-coated onto
morphologies through variation of microfluidic flow parameters, the silicon wafers and then heated to 95 °C for 60 min to remove residual
enabling facile processing control of colloidal nanostructures. SU-8 solvent. A photomask was placed over the film, and the device was
exposed to UV light for 180 s. After additional heating at 95 °C for 20
min, the device was submerged in SU-8 developer (Microchem) until all
’ EXPERIMENTAL SECTION
unexposed photoresist was removed. The reactor has a set channel depth
Materials. The polystyrene-block-poly(acrylic acid) sample used in of 150 μm and consists of a sinusoidal mixing channel 100 μm wide and
this study, PS(665)-b-PAA(68), was synthesized via anionic polymeri- 100 mm in length and a sinusoidal processing channel 200 μm wide and
zation of the associated polystyrene-block-poly(tert-butyl acrylate) block 740 mm in length.
copolymer, followed by hydrolysis of the ester block;47,48 numbers in For further stabilization of the bubble generation process, external
parentheses refer to number-average degrees of polymerization of each resistor chips were employed in series between the Ar gas tank and the
block. Dimethylformamide (DMF) (Aldrich, 99.9+%, HPLC grade, microfluidic chip. These resistors were 1000 mm long, 400 μm wide, and
H2O < 0.03%) and 1,4-dioxane Aldrich, 99.0%, reagent grade, H2O < 150 μm deep. In operation, these high pressure drop resistors serve to
0.05% were used as received without further purification. efficiently dampen the pressure fluctuations caused by the Ar gas tank
Critical Water Content Determination. In order to accurately and the bubble generation process itself. The resistors were designed in a
determine the critical water content (cwc) of PS(665)-b-PAA(68) for way that the total pressure drop in the resistors was at least 1 order of
the polymer concentration and solvent mixture applied in this study, magnitude higher than the pressure drop in the reaction channel.
static light scattering measurements were carried out. Light scattering Microfluidic chips were fabricated from poly(dimethylsiloxane)
experiments were performed on a Brookhaven Instruments photon (PDMS) using a SYLGARD 184 silicon elastomer kit (Dow Corning,
correlation spectrometer equipped with a BI-200SM goniometer, a BI- Midland, MI) with an elastomer base-to-curing agent ratio of 10:1. The
9000AT digital autocorrelator, and a Melles Griot He
Ne Laser elastomer and curing agent were mixed together and degassed in a
(632.8 nm) with a maximum power output of 75 mW. First, a stock vacuum chamber. The degassed PDMS was poured onto the negative
solution was prepared by adding 75/25 (w/w) DMF/dioxane solvent master in a Petri dish and then degassed again until all remaining air
mixture to a copolymer concentration of 0.66 wt %. The stock solution bubbles in the PDMS were removed. The PDMS was then heated at
was then filtered through a Teflon syringe membrane filter with a 85 °C until cured (∼50 min). The microfluidic chips were peeled off the
nominal pore size of 0.45 μm (National Scientific Company) into master, and holes were punched through its reservoirs to allow for the
precleaned scintillation vials. Next, the stock solution was diluted with insertion of tubing. A thin PDMS film was formed on a glass slide by
75/25 (w/w) DMF/dioxane solvent mixture filtered through two spin-coating and was permanently bonded to the base of the microfluidic
Teflon syringe filters with nominal pore size of 0.20 μm (National reactor after both components were exposed to oxygen plasma for 30 s.
Scientific Company) connected in series to give a polymer concentra- Flow Delivery and Control. Pressure-driven flow of liquids to the
tion of 0.33 wt %. Deionized water was next added dropwise to ∼6 g of inlets of the reactors was provided using 250 μL gastight syringes
the solution in successive 0.02
0.05 g quantities via a microsyringe (Hamilton, Reno, NV) mounted on syringe pumps (Harvard Apparatus,
equipped with a membrane filter (Life Sciences) with nominal pore size Holliston, MA). The microchip was connected to the syringes with
of 0.20 μm. After each addition of water, the solution was agitated with a 1/16th inch (OD) Teflon tubing (Scientific Products and Equipment,
vortexer to aid mixing. The scattered light intensity was recorded 15 min ON). Gas pressure was controlled by the Ar tank regulator as well as a
after vortexing to allow the solution to equilibrate. All measurements downstream regulator for fine adjustments (Johnson Controls Inc.). For
were carried out at a scattering angle of 90° and a temperature of 23 °C. gas flow, connections were joined using Teflon tubing of 1/16 in. (OD)
The cwc thus obtained (Figure S1, Supporting Information) was 3.2 ( and 100 μm (ID) (Upchurch Scientific, Oak Harbor, WA). The liquid
0.1 wt %. flow rate (Q liq) was programmed via the syringe pumps, and the gas flow
Off-Chip Micelle Preparation. Morphologies formed on-chip rate (Q gas) was varied by tuning the pressure regulator. Due to the
under different chemical and flow conditions were compared to off-chip compressible nature of the gas and the high gas/liquid interfacial
control experiments in which block copolymer micelles were prepared in tension, discrepancies arise between the nominal (programmed) and
the following manner. To three vials, each containing ∼5 mL of 0.33 wt % actual values of Q gas, Q gas/Q liq, and the total flow rate (Q total, also
PS(665)-b-PAA(68) in 75/25 (w/w) DMF/dioxane, deionized water denoted as Q for simplicity). Therefore, actual gas flow rates were
was added dropwise at a rate of 20 μL every 10 s with moderate magnetic calculated from the frequency of bubble formation and the average
stirring (600 rpm) until the various target water concentrations were volume of gas bubbles, determined from image analysis of the mean
reached (4.2, 5.2, and 7.2 wt %). The resulting micelles were then lengths of liquid and gas plugs, Lliq and Lgas, respectively, under a given
allowed to equilibrate for 2 weeks without stirring, followed by kinetic set of flow conditions. This method of flow calculation has been
trapping via transfer into ∼50 mL deionized water. The colloidal previously employed in the context of gas
liquid multiphase flow in
dispersions were finally dialyzed against deionized water to remove the microfluidic device.42,49,50 Actual flow parameters pertaining to the
residual organic solvent. gas
liquid segmented microfluidic reactor experiments are described in
In addition, to compare the effects of off-chip stirring-induced shear Table S1, Supporting Information. For all experiments, the relative gas-
with the effects of on-chip shear, we prepared micelles at 5.2 wt % water to-liquid flow ratio (Q gas/Q liq) was ∼1, and two different nominal total
in a manner similar to that described above, except for the following flow rates were investigated: Q = ∼5 μL/min (low flow rate case) and
differences. Once the target water concentration was reached, the ∼50 μL/min (high flow rate case) (see Table S1 for actual values).
micelles were stirred rapidly at 1200 rpm for a period of time equivalent On-Chip Micelle Preparation. For microfluidic self-assembly
to the on-chip residence time at low flow rate (280 s), followed by experiments, three separate liquid streams were combined at equal flow
immediate kinetic trapping into excess deionized water and dialysis. In rate to form gas-separated liquid plugs within the reactor: (1) a 1.0 wt %
Figure S2, Supporting Information, this additional off-chip control solution of PS(665)-b-PAA(68) single chains in 75/25 w/w DMF/
experiment is referred to as “off-chip (rapid stirring)”. dioxane; (2) a separator stream containing 75/25 w/w DMF/dioxane
Microfluidic Chip Fabrication. Negative masters were fabricated only; (3) a solution of 75/25 w/w DMF/dioxane containing 12.6, 15.6,
on high-quality silicon wafers (Silicon Quest International, Santa Clara, or 21.6 wt % deionized water. Combination of the three liquid streams
CA) using a negative photoresist, SU-8 100 (Microchem Inc.). Prior to yielded steady-state on-chip concentrations of 0.33 wt % polymer,
use, new silicon wafers were heated on a hot plate to at 200 °C for 20 min identical to the off-chip control experiments, and either 4.2, 5.2, or

18854 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864

Journal of the American Chemical Society ARTICLE

7.2 wt % water, 1.0, 2.0, and 4.0 wt % above the cwc (3.2 wt %), the sample. Second, the small number of fractions and very slow
respectively. Microfluidic flow conditions were selected as described in diffusion times for the large network micelles observed in many on-chip
the previous section. prepared samples did not allow them to be effectively sampled in typical
For off-chip analysis of micelle morphologies prepared within the collections of the DLS autocorrelation function.
reactor, micellar dispersions were generally collected from the chip Zeta Potential Measurements. Zeta potentials of various sam-
following the processing channel into vials containing a large excess ples were measured using a Zeta PALS Analyzer (Brookhaven Instru-
(∼250 μL) of deionized water, where the particles immediately became ments Corporation, Holtsville, NY). Zeta potential values were
kinetically frozen due to the high water content. For experiments computed from electrophoretic mobility values at 25 °C using the
tracking the off-chip relaxation of micellar morphologies formed on- Smoluchowski equation.
chip, micellar dispersions were collected from the chip following the
processing channel into empty vials; the solutions were then allowed to
sit off-chip without shear or stirring for various times, before kinetic ’ RESULTS AND DISCUSSION
trapping by dumping into excess deionized water. For both types of The specific amphiphilic block copolymer used for this study
micelle collection (immediate trapping or annealing), the solids content is polystyrene-block-poly(acrylic acid), PS(665)-b-PAA(68),
of the final mostly aqueous solution was ∼0.25 mg/mL. where numbers in brackets indicate number average degrees of
Flow visualization of microfluidic experiments and quantification of polymerization for the respective blocks. For all on-chip and off-
all the flow conditions was performed with an inverted fluorescence chip (control) experiments, solutions of 0.33 wt % PS(665)-b-
microscopy system operating in transmission mode. Optical images
PAA(68) in the solvent mixture 75/25 (w/w) N,N-dimethylfor-
of the reactor were capture using a CCD camera (AF6000 Orca,
mamide/1,4-dioxane (DMF/dioxane) serve as the initial state for
Hamamatsu, NJ) installed on the inverted microscope (DMI 6000B,
block copolymer self-assembly. Subsequent water addition to the
Leica, Allendale, NJ).
solution in the microfluidic device (on-chip) or in vials (off-chip)
TEM and Image Analysis. Transmission electron microscopy
increases the polarity of the solvent and decreases the solubility
(TEM) of various aqueous micellar dispersions was performed on a
Hitachi H-700 electron microscope, operating at an accelerating voltage
of the hydrophobic PS blocks until micelles with PS cores and
of 75 kV. For TEM imaging, a 10 μL drop of micellar dispersion was soluble PAA coronae are formed (Figure 1A). In both bulk and
deposited onto either a carbon-coated or carbon-coated Formvar 300 microfluidic environments, micelle formation is thermodynami-
mesh copper grid and then shadowed with Pt/Pd wire for imaging. For cally favorable above the critical water concentration (cwc),7
TEM imaging of on-chip prepared samples, the small quantity of dioxane which is determined to be 3.2 ( 0.1 wt % for the copolymer
remaining in the final solutions was found to dissolve Formvar sub- solution employed here (Figure S1). Therefore, to investigate
strates, and so carbon-coated grids without Formvar were required for the effect of the chemical environment on self-assembly, three
these samples. different water contents, 1, 2, and 4 wt % above the cwc, were
Analysis of micelle size distributions and statistics of both spherical selected for on-chip and off-chip experiments: that is, 4.2, 5.2, and
and nonspherical aggregates were carried on images randomly selected 7.2 wt % water in DMF/dioxane.
from various regions of the TEM grid. To assemble size histograms of Microfluidic self-assembly in the gas
liquid segmented re-
spherical micelle populations for each sample, the diameters of 150 actor is depicted in Figure 1B. A stream of the DMF/dioxane
spheres were measured with a caliper (Mastercraft, Boynton Beach, FL), solvent mixture containing PS(665)-b-PAA(68) copolymer is
from printouts of multiple regions of the TEM grid where the particles combined with a DMF/dioxane separator stream and a DMF/
appeared 40 000 to 50 000 times their actual size. dioxane/water stream; the concentrations of copolymer and
To quantify the percentage of nonspherical micelles in a given sample, water in the corresponding inlet streams are selected to give a
we determined their areal percentage relative to the total micelle area in cross-sectional average copolymer concentration of 0.33 wt %
TEM images. Assuming a constant density and constant height of and a cross-sectional average water concentration of 4.2, 5.2, or
different micelle morphologies on the TEM grids, the areal percentage 7.2 wt %. The streams are then joined with argon gas at the
of nonspherical micelles gives a reasonable estimate of their weight injector, which compartmentalizes the colaminar flow into liquid
percentage relative to total micelles in a given sample. The areal plugs. The resulting chaotic advection in the sinusoidal mixing
percentage of nonspherical micelles on a TEM grid was averaged from channel enables fast mixing of water with the copolymer solution
values determined from four to five randomly selected regions. When a (∼1 s),41,42,51,52 triggering micelle formation shortly after the
grid region was selected, a low-magnification (10k) image was first taken, injector.
followed by higher-magnification (20k) imaging of four quadrants of
A detailed schematic of the reactor is shown in Figure 1C. The
the larger area to allow for accurate image analysis. The higher magni-
reactor has a set channel depth of 150 μm and consists of a
fication images were processed with scientific image analysis software
sinusoidal mixing channel 100 μm wide and 100 mm in length,
(ImageProPlus v. 6.1, Media Cybernetics). Appropriate brightness,
contrast, and sharpness levels were selected for each image; after image
and a sinusoidal processing channel 200 μm wide and 740 mm in
binarization, threshold levels were set to allow particles to be counted length. Following initial formation, block copolymer micelles are
and analyzed by the software. By visual inspection, nonspherical objects exposed to continuous mixing as the liquid plugs proceed
were then selected and their areal percentage was calculated. For each through the mixing and processing channels, until the colloidal
sample, the process was repeated for two other TEM grids prepared dispersion is collected at the outlet for off-chip characterization
using the same micelle dispersion, in order to arrive at an average value by transmission electron microscopy (TEM). Unless otherwise
and experimental error (standard deviation of the three grids) for the stated, micelles formed on-chip are kinetically trapped immedi-
sample. The experimental errors on values of the areal percentage of ately upon exiting the chip by collection into a vial containing a
nonspherical micelles in various samples were ∼15
20%. large excess of deionized water; under these conditions, the
It was found that dynamic light scattering (DLS) analysis of many of amount of organic solvent in the PS phase becomes negligibly
the micelle populations prepared on-chip was not practical for two small, and micelle cores that were highly dynamic on-chip as a
reasons. First, the samples could not be effectively filtered to remove result of solvent swelling become suddenly vitrified, “freezing”
dust without also removing many of the larger nonspherical micelles in the on-chip colloidal morphologies for off-chip characterization.
18855 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

images highlight the broad range of morphologies that are

generated on-chip under these conditions, including linear and
looped cylinders, Y-junctions, spherical caps, bilayer sheets, and
dense networks of loops and cylinders. Several of the larger
micelles consist of hybrid morphologies described by abrupt
changes in local interfacial curvature. For example, the micellar
aggregate in Figure 2C possesses interconnected bilayer regions
with appendages of linear and looped cylinders, while Figure 2D
shows a flat bilayer sheet embedded within a network of looped
cylinders and Y-junctions.
On the basis of their uniformity and low polydispersity, we
expect that the spherical micelles formed off-chip (Figure 2A) are
close to thermodynamic equilibrium. Conversely, the various
nonspherical morphologies (Figure 2B
D) formed under iden-
tical chemical conditions but in the microfluidic device appear to
be kinetic products of the unique on-chip environment. Our
previous work on similar two-phase microfluidic reactors has
shown that recirculating flow in the liquid plugs gives rise to
localized regions of extremely high shear (∼104 to 105 s
1) in the
plug corners.41 Moreover, we have demonstrated that the
associated shear fields can induce both coalescence and breakup
of suspended, solvent-swollen polymeric particles, with shear-
induced coalescence dominating at low flow rate (∼5 μL/min)
and shear-induced breakup becoming increasingly important
at higher flow rates.42 In the current work, we attribute the
various nonspherical micelle morphologies formed at 5 μL/min
(Figure 2) to shear-induced coalescence of smaller, more ther-
modynamically stable, spherical micelles within the microfluidic
reactor.
The free energy pathway for such on-chip shear-induced mor-
Figure 1. PS-b-PAA self-assembly process and multiphase microfluidic phology transitions is represented in Scheme 1. First, spherical
reactor approach. (A) Schematic of the PS-b-PAA assembly process. Red micelles are generated shortly after the injector, as fast mixing
indicates hydrophobic PS blocks, while blue indicates hydrophilic PAA of water (<1 s) brings the average solvent composition above
blocks. (B) Cartoon representation of the self-assembly of PS-b-PAA the cwc (Scheme 1, State A). Next, as the liquid plugs travel
into micelles in the multiphase microfluidic reactor. The self-assembly is through the reactor, continuous recirculating flow moves
initiated by mixing of the three liquid stream contents within segmented micelles in and out of the high-shear corners, resulting in
plugs, which leads to an average cross-streamwater content above the shear-induced micelle coalescence to form larger micelles with
critical water content (cwc). (C) Schematic of the microfluidic reactor,
elevated free energies compared to the global equilibrium state
showing liquid and gas inlets at the injector, followed by the mixing
channel (represented in B); in the subsequent processing channel, the (Scheme 1, State B). Finally, internal chain rearrangements,
formed micelles experience the shear-induced collisions and morpho- driven by localized free energy lowering within coalesced
logical transformations discussed in the text, which are trapped by micelles, give rise to the various nonspherical morphologies
collection into pure water after the outlet. The inset of C shows a select (Scheme 1, State C).
optical microscopy image of the reactor under stable operation. The Shear-induced aggregation of solid colloidal particles (or
white scale bar indicates 500 μm. coalescence of liquid drops) is a well-known phenomenon
occurring in both natural and industrial systems, with examples
For all experiments, the gas-to-liquid flow ratio is ∼1, and ranging from the deterioration of dispersed pigments to the
two different nominal total flow rates are investigated: Q = formation of strong silk fibers within a spider’s spinneret.54
∼5 μL/min (low flow rate case) and ∼50 μL/min (high flow rate Recent theoretical work by Zaccone and co-workers describes
case).53 An optical image of part of the reactor (Figure 1C, inset) shear-induced aggregation as an activated-rate process, using an
demonstrates typical stable operation in which uniform and Arrhenius-type rate equation to show that shear both increases
regularly spaced gas bubbles are produced. the particle collision rate and also supplies energy to overcome
The dramatic effect of the microfluidic environment on repulsive interparticle potentials, thus increasing the fraction of
micelle morphologies produced by block copolymer self-assembly adhesive collisions.55 In general, the relative effects of shear and
is shown in Figure 2. In the off-chip control experiment, water Brownian motion on aggregation rates of colloidal particles can
is added dropwise to a solution of PS(665)-b-PAA(68) up to be described by the dimensionless Peclet number:
5.2 wt %, and the resulting dispersion is allowed to equilibrate for :
2 weeks, generating exclusively spherical micelles (Figure 2A, γa2
Pe ¼ ð1Þ
mean diameter, d = 35 nm). In contrast, block copolymer self- D0
assembly at the same water content in the microfluidic device
operating at low flow rate (Q = ∼5 μL/min) gives a mixture where γ_ is the shear rate, a is the particle size and D0 is the particle
of spherical micelles (d = 41 nm) and several higher order, diffusion coefficient.56 For low Peclet numbers (Pe , 1), thermal
nonspherical morphologies (Figure 2B
D). The representative diffusion has a dominant effect on particle aggregation, whereas
18856 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

Figure 2. TEM images comparing off-chip (A) and on-chip, Q = ∼5 μL/min (B
D), morphologies of PS-b-PAA micelles formed in DMF/dioxane
mixtures containing 5.2 wt % water. Only the equilibrium spherical morphology was formed off-chip (A). However, the on-chip experiment yielded a
mixture of morphologies in addition to spheres, including linear and looped cylinders (B, C, and D), Y-junctions (B and D), spherical caps (B), bilayer
sheets (C and D), and networks (D). All scale bars = 200 nm, except the higher-magnification inset to A (scale bar = 50 nm).

shear forces have a stronger influence on aggregation rates as water within the PAA corona.60 This means that the PAA
Peclet numbers increase.57 chains will be partially ionized and negatively charged, with the
From eq 1, the initially formed spherical micelles (a = extent of ionization increasing at higher water contents; this
∼40 nm) in the high-shear corners of the liquid plugs at low was confirmed by off-chip zeta-potential measurements of
flow rate (γ_ = ∼104 s
1) are described by a Peclet number of Pe = PAA-b-PAA micelles at the same solvent conditions as on-chip
∼10.58 Although the specific dependence of shear-induced experiments, which gave ζ =
22 ( 4,
30 ( 2, and
33 (
aggregation on Pe depends on the nature of the interparticle 2 mV in DMF/dioxane at 4.2, 5.2, and 7.2 wt % water,
interactions,56 Pe numbers on the order of ∼10 have been found respectively.
to give rise to shear-induced aggregation in related systems.55
57 To estimate the contribution of intercoronal repulsion to Ea1,
It is therefore reasonable to propose that micelles will experience we employ a recent model by Jusufi et al. describing repulsive
multiple shear-induced collisions as they circulate through interactions between spherical polyelectrolyte brushes.59 Ac-
the high-shear hot spots of the plug corners, with some of cording to this model, the confinement of counterions within
these collisions being sufficiently energetic to overcome Ea1 the brush layers results in negligible electrostatic interparticle
(Scheme 1) and activate micelle coalescence.55 The collision of forces. As well, it is assumed that chains in the approaching
two primary micelles, represented in Scheme 1, is only one of brushes retract from each other (or “dent”) rather than inter-
many statistical possibilities, which will also include collisions digitating. The dominant term of the resulting electrosteric pair
between one or more larger micelles (with higher Peclet num- potential is attributed to the decreased entropy of counterions in
bers, eq 1) that have experienced at least one previous instance of the overlapping polyelectrolyte brushes. The analytical form of
collision
coalescence. their result is a pair potential that increases from zero as
A key component of the activation energy Ea1 is the repulsion approaching brushes begin to overlap and reaches a maximum
between the overlapping coronae of colliding micelles (Scheme 1).59 value of 0.2kT per unit counterion when the cores come into
Even at the relatively low water contents in DMF/dioxane contact. From this result, we take 0.2kT  (Q /|e|) as the
employed on-chip (∼4
7 wt %), PS-b-PAA micelles will exhibit electrosteric contribution to Ea1, where Q is the effective charge
polyelectrolyte character, due to preferential accumulation of of counterions in the corona and e is the elementary charge.
18857 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

Scheme 1. Proposed Energy Diagram for On-Chip Shear- coalescence becomes equal to the collision rate, when γ_ . γ*:
_
Induced Morphological Transitions
: Ea1
γ ¼ ð2Þ
6παμa3

.In this expression, α is a numerical coefficient describing the flow

(e.g., α = 1/3π for simple shear) and μ is the dynamic viscosity.
Using Ea1 = 100kT, α = 1/3π, μ = 0.001 Pa s, and a = 40 nm, we
obtain γ*_ = 6  106 s
1 for primary spherical PS-b-PAA micelles
in DMF/water with electrosteric repulsion. Because in the high-
shear hot spots on chip γ* _ . γ_ = ∼104 s
1, we conclude that the
magnitude of Ea1 will strongly influence the kinetics, such that the
rate of shear-induced coalescence is expected to increase as Ea1
decreases with decreases in coronal charge or core viscosity.55
As the average micelle size increases through statistical colli-
sion-coalescence events, increased PS chain stretching stores
accumulated shear energy in micelles with elevated free energies
(Scheme 1, State B). Outside of the high-shear regions, coalesced
micelles in State B will therefore have a thermodynamic tendency
to return to their lowest free energy state via micelle fission
(Scheme 1). However, if a sufficiently high Ea1 results in a fission
rate that is comparable or slow compared to the rate of shear-
induced coalescence, then on-chip relaxation to global equilibrium
will be precluded. Under this kinetic constraint, a significant
number of micelles in State B will lower their free energy via the
alternative pathway of intramicellar chain rearrangements, lead-
ing to localized morphology transitions (State C). The energy
barrier Ea2 is expected to be a function of the viscous resistance to
intramicellar chain diffusion within the PS cores; at low water
For spherical micelle cores with a = ∼40 nm, an aggregation contents (4
5 wt %), the PS cores will be highly swollen
number of Z = ∼250 is determined, or ∼2  104 COOH units with solvent, with estimated core polymer concentrations of
per micelle. Assuming a maximum degree of ionization of ∼10% ∼50 wt %.8 Using literature values for the activation energy of
(calculated from a pKa of 4.25 in water),60 and two-thirds viscous flow for PS
toluene solutions of similar concentra-
condensed counterions,59 we obtain an effective charge of tions,61 we estimate Ea2 to be ∼10kT, 1 order of magnitude
counterions in the corona of Q = ∼700 |e|, which yields an Ea1 lower than Ea1. This explains why the intramicellar pathway to
value on the order of ∼100kT if electrosteric effects dominate the the observed nonspherical morphologies is kinetically favored
coalescence energy barrier. over a return to global equilibrium. Therefore, as a combined
Another potential contribution to Ea1 is the viscosity of the result of the statistical nature of shear-induced collisions and the
PS micelle cores, which will govern the rate of core coalescence kinetic constraints of relaxation processes in the microfluidic
if the electrosteric barrier described above is overcome. It is device, we obtain a potpourri of spherical and nonspherical
well established that block copolymer intermicellar and in- morphologies that are not globally equilibrated but nonetheless
tramicellar chain dynamics are a strong function of the water represent a tendency of micelles to locally minimize their internal
content in polar organic solvents; at low water contents, the free energy.
micelle cores are highly swollen by the organic solvent, Although the details of these on-chip morphological transfor-
becoming increasingly viscous until eventual kinetic freezing mations are currently unknown, copolymer chains within a
as the water content increases.8,10 For PS-b-PAA in DMF, fast coalesced micelle are assumed to redistribute following micro-
micelle equilibration kinetics on experimental time scales for phase separation principles, with internal curvature optimized to
water contents less than ∼6.5% have been found, suggesting reduce chain stretching within the constraints of high packing
highly swollen and dynamic PS blocks in this range of solvent densities required to lower the interfacial tension.1,2,13,34 In
composition;10 based on this result, and considering that general, micelle morphologies will change from spheres to
dioxane is an even better solvent for the PS core than DMF,8 cylinders to bilayer structures (in order of decreasing internal
it is reasonable to conclude that core viscosities do not curvature), as a means of relieving stretching of core-forming
contribute significantly to Ea1 in DMF/water mixtures at the blocks in response to chemical variables that increase the chain
two lowest water contents of 4.2 and 5.2 wt %. However, as the packing density at the interface.1,2 In the present case, it is shear-
water content increases, the solvent content in the core, along induced coalescence, rather than chemical forces, that locally
with the chain dynamics, will drop off significantly; we there- increase aggregation numbers, triggering localized transitions
fore cannot rule out a considerable, and possibly even domi- from spheres to mainly cylindrical structures in order to lower
nant, contribution of core viscosity to the coalescence rate at internal curvature and reduce PS stretching. However, despite
the highest water content (7.2 wt %). the predominance of cylinders under the current chemical and
From the Arrhenius expression for the rate of shear-activated flow conditions, we point out that other nonspherical morphol-
_ can be defined such that the
coalescence,55 a critical shear rate γ* ogies with even lower internal curvature are also observed,
activation barrier becomes vanishingly small, and the rate of suggesting significant variability in local conditions due to the
18858 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

Table 1. Mean Spherical Micelle Diameters under Various

On-Chip and Off-Chip Preparation Conditions
conditions 4.2 wt % water 5.2 wt % water 7.2 wt % water

Off-Chip
29 nm 35 nm 39 nm
On-Chip, ∼5 μL/min
t=0 41 nma 41 nma 33 nm
t=2h 41 nma 36 nma

t = 10 h 45 nma 36 nma

t = 48 h 49 nm a
35 nma

t = 7 days 49 nm 35 nma

t = 28 days 49 nm 35 nm

On-Chip, ∼50 μL/min
37 nm 32 nm 30 nm
a
Indicates sample containing nonspherical micelles, in addition to the
spherical micelles measured here.

resulting spherical micelles are similar in size (d = 37 nm) to

micelles formed off-chip at 5.2 wt % water without vigorous
stirring (d = 35 nm, Figure S2). We conclude that nanoscale
velocity gradients in the plug corners are germane to increasing
the rate of on-chip collision
coalescence events relative to
normal diffusive motion in the bulk, thus enabling the morphol-
ogy transformations described above.
Figure 3 describes the effect of chemical conditions on the on-
Figure 3. The effect of water content on on-chip PS-b-PAA micelle
chip formation of flow-directed block copolymer micelle mor-
morphologies (A, C, and E) and spherical micelle size distributions phologies at low flow rate. In Figure 3A,C,D, representative
(B, D, and F) formed in DMF/dioxane mixtures, Q = ∼5 μL/min. A, C, micelles are shown for solvent mixtures with 4.2, 5.2, and 7.2 wt %
and D show TEM images of representative micelles formed at 4.2, 5.2, water, with associated spherical micelle size distributions shown
and 7.2 wt % water, respectively; B, D, and F show size histograms of in Figure 3B,D,F. Not shown in the figure are the results of the
spherical micelle populations formed at the corresponding water con- associated off-chip control experiments, which give rise to uni-
tents. All scale bars = 200 nm. form and low-polydispersity spherical micelles for all three water
contents, with monotonically increasing mean sphere sizes of
statistical nature of collision-coalescence events. For instance, the 29 nm, 35 nm, and 39 nm for 4.2, 5.2, and 7.2 wt % water,
observed hybrid morphologies discussed earlier (e.g., Figure 2, C respectively (Table 1). Comparing 4.2 and 5.2 wt % water on-
and D) are explained by intraparticle morphology transitions chip (Figure 3A,C), both images reveal a mixture of spheres and
(e.g., from cylinders to bilayers) that become kinetically trapped nonspherical morphologies, indicating that shear-induced mor-
during off-chip collection. phological transformations are operative at both water contents.
Recently, Winnik and co-workers showed that a sphere-to- In addition, both samples show spherical micelle populations
wormlike network transition could be induced in polystyrene- with the same mean size (d = 41 nm), larger than the associated
block-poly(4-vinylpyridine) (PS-b-P4VP) block copolymer mi- off-chip equilibrium sizes (d = 29 and 35 nm for 4.2 and 5.2 wt %
celles with PS cores and P4VP coronae by conventional rapid water, respectively), although with very different distribution
magnetic stirring at 1200 rpm but only in the presence of in CdSe shapes and widths (Figure 3B,D). We also find that the higher
quantum dots, which were proposed to form bridging inter- water content results in significantly larger network structures
actions between coronae of assembling spherical micelles.29 (Figure S3), as well as a higher areal percentage of nonspherical
Therefore, to confirm the important role of the extremely strong micelles relative to the total micelle area (areal percentage = 5%
shear fields enabled by the on-chip environment in the current and 24% at 4.2 and 5.2 wt % water, respectively).62 This indicates
system,41 we investigate the effect of micelle formation under off- that a significantly larger percentage of spherical micelles are
chip vigorous magnetic stirring conditions similar to those converted into nonspherical nanostructures by the on-chip
employed in ref 34. In this experiment, micelles are formed by environment at 5.2 wt % compared to 4.2 wt % water. Interest-
dropwise addition of water to 5.2 wt % and then stirred at ingly, the opposite trend is found upon further increasing the
1200 rpm for 280 s (equivalent to the on-chip residence time at water content to 7.2 wt % (Figure 3E), which results in uniform
low flow rate). Although the solution mixing will be much spherical micelles with mean size d = 33 nm (Figure 3F), smaller
improved, the maximum shear rates in this experiment are on than the associated equilibrium value (d = 39 nm), with no
the order of 1 s
1, which are negligible compared to local maxima observable nonspherical micelles.
experienced on-chip. In contrast to on-chip self-assembly under The influence of water content on the on-chip morphologies
the same chemical conditions, no nonspherical micellar struc- can be understood in terms of its effect on activation barriers in
tures are formed in these more conventional shear fields, and the Scheme 1. The activation energy for the coalescence/fission
18859 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

process, Ea1, is a function of intercoronal electrosteric repulsion the liquid plug volume is exposed to the high shear regions during
and core viscosity for two colliding micelles, which represents the the initial mixing/micellization process.65 The large percentages
transition state between States A and B.63 With increasing water of nonspherical structures that we observe under certain condi-
content, intercoronal repulsion and core viscosity will both tions (up to 25% of the total micelle volume) cannot, therefore,
increase, as the PAA chains become increasingly ionized and be accounted for by micellization under shear. While it is not
the solvent quality for the PS core decreases.10 Both effects will possible to completely shelter the bulk liquid from high shear
therefore contribute to an increase in Ea1, and the latter effect will during the short mixing process, the rapid mixing achieved with
also contribute to an increase in Ea2. The increase in electrosteric the multiphase microfluidic system effectively separates the
interactions will dominate at the two lowest water contents (4.2 mixing function, which initiates micelle formation, from the
and 5.2 wt %), where the electrosteric barrier for coalescence is downstream shear processing function, which induces the mor-
∼10 greater than the barrier for viscous flow within the core. phological transformation.
However, the increase in core viscosity is expected to play a more A second mechanism that merits consideration is one in
significant role at 7.2 wt % water, based on the demonstrated which the directionality of shear forces templates the formation
sharp decrease in PS-b-PAA chain dynamics for water contents of specific micelle morphologies. For example, cylindrical mi-
>6.5 wt %.10 celles could be formed via the one-dimensional assembly of
For an intermediate value of Ea1, sufficiently low for shear- spherical micelles along the direction of shear. However, we
induced coalescence to be activated but sufficiently high to eliminate such a “shear-templating” mechanism based on the
significantly inhibit the reverse fission process, a large number presence of other, nondirectional, morphologies such as bilayer
of high-energy micelles will be trapped on the right-hand-side of sheets (Figure 2, C and D) which also form as a result of shear
the coalescence/fission activation barrier. This situation is con- exposure. Rather, as described above, our results strongly
sistent with the bimodal nature of the spherical micelle distribu- support a mechanism in which the role of on-chip shear forces
tion at 5.2 wt % (Figure 3D), in which two distinct populations of is simply to increase the rate of micelle coalescence in a statistical
smaller and larger particles correspond to micelles in States A and manner, such that colloids with aggregation numbers well above
B, respectively. In contrast, when Ea1 is lowered by decreasing the the mean equilibrium value are generated. The subsequent
water content (e.g., to 4.2 wt % water), a faster fission rate leads relaxation of these aggregates via intramicellar pathways is then
to a more dynamic exchange between States A and B, resulting in guided by internal molecular forces (and not shear forces),
a single averaged population of spherical micelles (Figure 3B). As ultimately forming the micelle morphology that allows the
well, the larger barrier Ea1 at 5.2 wt % compared to 4.2 wt % forces greatest decrease in local free energy relative to the initial
a greater number of micelles in State B to follow the relatively fast shear-induced aggregation state.
intraparticle relaxation pathway through Ea2, explaining the larger Several of the on-chip morphologies formed at low flow rate,
quantity of nonspherical micelles in the former case. Conversely, including looped cylinders, Y-junctions, and networks of linear
when the water content is increased to the point where Ea1 is too and looped cylinders, are very reminiscent of nonergodic mi-
high for micelle coalescence to be activated (e.g., to 7.2 wt % celles of poly(1,2-butadiene)-block-poly(ethylene oxide) (PB-b-
water), the formation of nonspherical morphologies is precluded, PEO) in aqueous solution described previously by Bates and co-
and only spherical micelles are observed (Figure 3E). workers.13 Although micelles in that work were not formed in the
The effect of flow rate on microfluidic block copolymer self- presence of strong external shear fields, the morphological
assembly was also determined, by investigating micelles prepared similarities with our system are consistent with on-chip micelles
at the same three water contents but at high flow rate (∼50 μL/ being nonergodic on the time scale of our microfluidic experi-
min, Figure S4, Supporting Information). In contrast to the low ments. In the work of Bates, complex micelles with unusual
flow rate case, only unimodal populations of spherical micelles, mirror symmetry were attributed to network fragmentation by
with no additional morphologies, are observed for all three water stirring or sonication followed by intraparticle redistribution of
contents. According to our previous results on kinetic polymeric molecules to balance the free energy, under conditions where
particles formed in similar reactors, the on-chip competition interparticle exchange of material could not occur on experi-
between shear-induced particle coalescence and shear-induced mental time scales.13 Interestingly, in addition to cylinder net-
particle breakup favors coalescence at low flow rates and breakup work fragments of various sizes formed on-chip (e.g., Figure 4A),
as the flow rate increases.42 Therefore, for the 4.2 and 5.2 wt % we find a number of compelling complex micelles with mirror
water contents, when the flow rates are increased to ∼50 μL/min, symmetry (Figure 4B
E). This further highlights the impor-
the shear-induced coalescence that gives rise to nonspherical tance of intraparticle chain rearrangements in the on-chip for-
morphologies at ∼5 μL/min (Figure 3, parts A and C, res- mation of the observed micelle morphologies.
pectively) is countered by the dominant mechanism of shear- In contrast to PB-b-PEO micelles in water, which showed
induced particle breakup, such that cylinders and other higher- nonergodic behavior over a time scale of months to years, the off-
aggregation structures cannot form. chip relaxation of PS-b-PAA micelles formed on-chip can be
On the basis of the current results, we believe that two monitored over a time scale of days to weeks (Figure 5). For
alternative mechanisms for the on-chip morphology transforma- relaxation studies, micellar dispersions are collected from the
tions can be ruled out. The first of these is a shear-induced chip and allowed to sit for various times (t) before kinetic
micellization picture, described in the theoretical work of Jones trapping by dumping into excess water; t = 0 represents samples
et al.,64 in which micellized states are influenced by shear forces at collected with immediate off-chip trapping, giving a mixture of
the stage of unimer self-assembly. To eliminate this possible spherical and nonspherical morphologies as discussed above
mechanism, we consider that only a very small percentage of (Figure 5, A and C, for 4.2 and 5.2 wt % water, respectively).
micelles will initially form under significant shear, due to the As shown in Figure 5, B and D, after sufficient off-chip time for
strong localization of high-shear regions in the plug corners. both water contents, t = 7 days for 4.2 wt % (Figure 5B) and t = 28
Specifically, a conservative calculation shows that less than 2% of days for 5.2 wt % (Figure 5D), the nonspherical micelles
18860 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

Figure 4. Further examples of PS-b-PAA micellar structures formed on-chip, Q = ∼5 μL/min, in DMF/dioxane mixtures containing 4.2 wt % water
(A,B, D, and E) or 5.2 wt % water (C). TEM images B
E highlight the prominent occurrence of nonspherical micelles with mirror image symmetry from
on-chip self-assembly. All scale bars = 200 nm.

disappear completely, leaving only uniform populations of Mean decay times for each fit were calculated using
spheres. Because spherical micelles have been shown to be the    
equilibrium morphology at both of these water contents, the a b
τ̅ ¼ τ1 þ τ2 ð4Þ
disappearance of nonspherical micelles is consistent with the a þ b a þ b
relaxation of shear-induced structures toward global equilibrium,
following their transfer to off-chip quiescent conditions. Note giving 26 s for the 4.2 wt % water sample and 5200 s (∼2 h) for
that the times required for the disappearance of cylindrical the 5.2 wt % water sample.
structures, on the time scale of days, are extremely long compared Along with different relaxation rates, a more detailed analysis
to the on-chip residence time (280 s); this explains why signi- of TEM images provides evidence for different off-chip relaxation
ficant on-chip relaxation does not occur, such that shear- mechanisms following on-chip micelle formation at 4.2 and 5.2
induced morphologies could be kinetically trapped off-chip for wt % water. The inset to Figure 5E shows an example of
observation. prominent aggregates observed between t = 10 min and t =
The off-chip relaxation kinetics of nonspherical micelles can 120 min in dispersions formed at 4.2 wt % water, which appear to
be quantified by tracking the time-dependent decrease of their be transition structures formed by cylinder networks breaking
relative area for 4.2 and 5.2 wt % water dispersions. Because these into spheres in the manner of Rayleigh instabilities in liquid
samples show different initial percentages of nonspherical mi- strands.66 The absence of similar structures during relaxation of
celles, as discussed earlier, the two decay functions are compared the 5.2 wt % water sample suggests that cylinder breakup is not a
by normalizing each data point with respect to the initial areal prominent relaxation mechanism in that case, presumably due to
percentage value at t = 0. Figure 5E shows the resulting decay stronger electrosteric repulsion between proximal spheres. This
plots for 4.2 and 5.2 wt % water cases, indicating slower off-chip difference implies that the off-chip transition between nonsphe-
disappearance of nonspherical structures at higher water content. rical and spherical micelles at 5.2 wt % water occurs via the
We found that neither data set could be fit to a single exponential alternative relaxation mechanism of unimer exchange, with larger
decay, although both plots fit reasonably well to biexponential aggregates progressively releasing block copolymer chains into
decay functions: solution which then reassemble into more stable spheres.
Additional support for these different relaxation mechanisms
rel area nonspherical micelles ¼ ae
t=τ1 þ be
t=τ2 ð3Þ comes from following the evolution of spherical micelle size
distributions as a function off-chip time (Figure 6). Size distribu-
The individual fast and slow decay times from the fit, τ1 and tions for 4.2 and 5.2 wt % water are shown on the left-hand and
τ2, are probably not physically meaningful, although they char- right-hand sides of the figure, respectively, for increasing off-chip
acterize the distribution of decay times in the collected samples. times, t, moving down each column. The equilibrium (off-chip)
18861 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

Figure 6. Evolution of spherical micelle size distributions as a function

off-chip relaxation time, t, for the two on-chip samples (4.2 and 5.2 wt %
Figure 5. Off-chip relaxation kinetics of nonspherical micelles formed water) described in Figure 5. The expected equilibrium mean size of
on-chip (Q = ∼5 μL/min) at two different water contents. A and C show spherical micelles under the same chemical conditions, based on off-chip
TEM images of immediately trapped (t = 0) mixtures of spherical and micellization experiments at 4.2 or 5.2 wt % water followed by two weeks
nonspherical micelles formed at 4.2 and 5.2 wt % water, respectively. equilibration (Table 1), is indicated as a dashed line in the histograms.
TEM images in B and D indicate that only spherical micelles remain at
both water contents after sufficient off-chip relaxation time (t = 7 and 28 In contrast, the size distributions of spherical micelles show
days for 4.2 and 5.2 wt % water, respectively). In E, the relative area of
very different off-chip evolution at 5.2 wt % water. In this case,
nonspherical micelles is plotted vs t for 4.2 wt % water (blue triangles)
and 5.2 wt % water (red circles), with mean decay times of 26 and 5200 s, the bimodal distribution discussed previously (t = 0) rapidly
respectively, determined from fits to biexponential decay functions converges to a unimodal sphere population (t = 2 h, d = 36 nm)
(solid blue and red curves). The inset to E shows an example of centered close to the equilibrium size (d = 35 nm), which then
transition structures suggesting break-up of cylinders into spheres persists for the remainder of the monitored time off-chip. This
during off-chip relaxation at 4.2 wt % water. All scale bars = 200 nm. trend corroborates the off-chip relaxation mechanism of unimer
exchange at this water content. By this pathway, the cylinder
networks disappear more gradually than by cylinder breakup
mean sphere sizes for both water contents are indicated as a (Figure 5), although in a manner that allows global equilibrium to
dashed line in the histograms. For 4.2 wt % water, no changes in be reached on a shorter time scale.
the position and shape of the distribution are observed in the first
2 h off-chip (d = 41 nm), although by 10 h a clear shift to larger
spherical micelles has occurred (d = 45 nm), away from the ’ CONCLUSIONS
equilibrium sphere size (d = 29 nm). This unexpected shift can be We have demonstrated that the solution self-assembly of
explained by the process of cylinder breakup and the resulting amphiphilic block copolymers in a gas
liquid multiphase micro-
generation of daughter spheres, whose sizes are governed by local fluidic reactor provides flow-directed variation of micelle mor-
wave instabilities that develop in the cylinder during relaxation phologies due to the unique on-chip shear environment. Under
rather than by a global free energy minimum. Following the near- chemical conditions that form equilibrium spherical micelles off-
complete disappearance of nonspherical micelles within ∼1 day chip, a range of nonspherical nanostructures, including bilayer
at 4.2 wt % water (Figure 5), the resulting spheres show no lamellae and networks of cylinders, can be generated on-chip.
evidence of approaching their equilibrium size even after 28 days The resulting nanostructures are varied using a combination of
(Figure 6). We conclude that the relatively fast off-chip relaxation chemical (bottom-up) and flow (top-down) parameters, and can
process of cylinder breakup leads to nonergodic spherical be kinetically trapped off-chip by collection into excess water. A
micelles kinetically trapped within a local free energy minimum. micelle formation mechanism of shear-induced collision
coalescence
18862 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864
Journal of the American Chemical Society ARTICLE

followed by intraparticle chain rearrangements is proposed for (19) Li, Z.; Hillmyer, M. A.; Lodge, T. P. Nano Lett. 2006, 6,
microfluidic self-assembly. Finally, the off-chip micelle relaxation 1245–1249.
processes are analyzed under different chemical conditions. Top- (20) Larue, I.; Adam, M.; Pitsikalis, M.; Hadjichristidis, N.; Rubinstein,
down directed molecular self-assembly in microfluidic reactors opens M.; Sheiko, S. S. Macromolecules 2006, 39, 309–314.
up new opportunities for processing control of functional colloidal (21) Allen, C.; Maysinger, D.; Eisenberg, A. Colloids Surf., B 1999,
16, 3–27.
nanostructures for a wide range of applications ranging from imaging (22) Geng, Y.; Dalhaimer, P.; Cai, S.; Tsai, R.; Tewari, M.; Minko,
to drug delivery. T.; Discher, D. E. Nat. Nanotechnol. 2007, 2, 249–255.
(23) Gonzalez, D. C.; Savariar, E. N.; Thayumanavan, S. J. Am. Chem.
’ ASSOCIATED CONTENT Soc. 2009, 131, 7708–7716.
(24) Torchilin, V. P. Adv. Drug Delivery Rev. 2002, 54, 235–252.
bS Supporting Information. Data for critical water content (25) Nasongkla, N.; Bey, E.; Ren, J. M.; Ai, H.; Khemtong, C.; Guthi,
determination; TEM images of micelles prepared off-chip with J. S.; Chin, S. F.; D., S. A.; Boothman, D. A.; Gao, J. M. Nano Lett. 2006,
different stir rates; additional TEM images of nonspherical 6, 2427–2430.
micelles generated on-chip at low flow rate and different water (26) Wu, W. C.; Chen, C. Y.; Tian, Y. Q.; Jang, S. H.; Hong, Y. N.;
contents; TEM images and particle size distributions of spherical Liu, Y.; Hu, R. R.; Tang, B. Z.; Lee, Y. T.; Chen, C. T.; C., C. W.; Jen,
A. K. Y. Adv. Funct. Mater. 2010, 20, 1413–1423.
micelles generated on-chip at high flow and different water (27) Rockford, L.; Mochrie, S. G. J.; Russell, T. P. Macromolecules
contents; actual measured flow data for various runs described 2001, 34, 1487–1492.
in the text; relative standard deviations for data in Table 1. This (28) Kim, S. O.; Solak, H. H.; Stoykovich, M. P.; Ferrier, N. J.;
information in available free of charge via the Internet at de Pablo, J. J.; Nealey, P. F. Nature 2003, 424, 411–414.
http://pubs.acs.org/. (29) Segalman, R. A.; Hexemer, A.; Kramer, E. J. Macromolecules
2003, 36, 6831–6839.
’ AUTHOR INFORMATION (30) Cheng, J. Y.; Mayes, A. M.; Ross, C. A. Nat. Mater. 2004, 3,
823–828.
Corresponding Author (31) Stoykovich, M. P.; Muller, M.; Kim, S. O.; Solak, H. H.;
mmoffitt@uvic.ca Edwards, E. W.; De Pablo, J. J.; Nealey, P. F. Science 2005, 308, 1442–
1446.
(32) Cheng, J. Y.; Rettner, C. T.; Sanders, D. P.; Kim, H. C.;
’ ACKNOWLEDGMENT Hinsberg, W. D. Adv. Mater. 2008, 20, 3155–3158.
(33) Wang, J. F.; Chen, Q. J.; Zhao, H.; Ming, T.; Wu, C. J. Am.
The authors are grateful for the financial support by the Chem. Soc. 2009, 131, 16650–16651.
Natural Sciences and Engineering Research Council (NSERC) (34) Zhang, M.; Wang, M. F.; He, S.; Qian, J. S.; Saffari, A.; Lee, A.;
of Canada, through research grants to D.S., and MGM Infra- Kumar, S.; Hassan, Y.; Guenther, A.; Scholes, G.; Winnik, M. A.
structure funding from the Canada Foundation for Innovation Macromolecules 2010, 43, 5066–5074.
(CFI). British Columbia Knowledge Development Fund (BCKDF) (35) Nie, Z.; Xu, S.; Seo, M.; Lewis, P. C.; Kumacheva, E. J. Am.
is also gratefully acknowledged, along with a Collaborative Chem. Soc. 2005, 127, 8058–8063.
Research Stipend from the Centre for Advanced Materials and (36) Xu, S. Q.; Nie, Z. H.; Seo, M.; Lewis, P.; Kumacheva, E.; Stone,
Related Technologies (CAMTEC, UVic). H. A.; Garstecki, P.; Weibel, D. B.; Gitlin, I.; Whitesides, G. M. Angew.
Chem., Int. Ed. 2005, 44, 724–728.
(37) Seo, M.; Nie, Z.; Xu, S.; Mok, M.; Lewis, P. C.; Graham, R.;
’ REFERENCES Kumacheva, E. Langmuir 2005, 21, 11614–11622.
(1) Zhang, L.; Eisenberg, A. Science 1995, 268, 1728–1731. (38) Karnik, R.; Gu, F.; Basto, P.; Cannizzaro, C.; Dean, L.; Kyei-
(2) Zhang, L.; Eisenberg, A. J. Am. Chem. Soc. 1996, 118, 3168–3181. Manu, W.; Langer, R.; Forokhzad, O. C. Nano Lett. 2008, 8, 2906–2912.
(3) Yu, K.; Zhang, L.; Eisenberg, A. Langmuir 1996, 12, 5980–5984. (39) Anderson, D. G.; Xu, Q. B.; Hashimoto, M.; Dang, T. T.;
(4) Zhang, L.; Yu, K.; Eisenberg, A. Science 1996, 272, 1777–1779. Hoare, Y.; Kohane, D. S.; Whitesides, G. M.; Langer, R. Small 2009,
(5) Shen, H.; Zhang, L.; Eisenberg, A. J. Phys. Chem. B 1997, 24, 5, 1575–1581.
4697–4708. (40) Schabas, G.; Yusuf, H.; Moffitt, M. G.; Sinton, D. Langmuir
(6) Yu, Y.; Eisenberg, A. J. Am. Chem. Soc. 1997, 119, 8383–8384. 2008, 24, 637–643.
(7) Zhang, L.; Shen, H.; Eisenberg, A. Macromolecules 1997, 30, (41) Schabas, G.; Wang, C.-W.; Oskooei, A.; Yusuf, H.; Moffitt,
1001–1011. M. G.; Sinton, D. Langmuir 2008, 24, 10596–10603.
(8) Yu, Y.; Zhang, L.; Eisenberg, A. Macromolecules 1998, 31, (42) Wang, C.-W.; Oskooei, S. A. K.; Sinton, D.; Moffitt, M. G.
1144–1154. Langmuir 2010, 26, 716–723.
(9) Discher, B. M.; Won, Y.-Y.; Ege, D. S.; Lee, J. C.-M.; Bates, F. S.; (43) Brown, L.; McArthur, S. L.; Wright, P. C.; Lewis, A.; Battaglia,
Discher, D. E.; Hammer, D. A. Science 1999, 284, 1143–1146. G. Lab Chip 2010, 10, 1922–1928.
(10) Zhang, L.; Eisenberg, A. Macromolecules 1999, 32, 2239–2249. (44) Jahn, A.; Vreeland, W. N.; Gaitan, M.; Locascio, L. E. J. Am.
(11) Shen, H.; Eisenberg, A. J. Phys. Chem. B 1999, 103, 9473–9487. Chem. Soc. 2004, 126, 2674–2675.
(12) Discher, D. E.; Eisenberg, A. Science 2002, 297, 967–973. (45) Steinbacher, J. L.; Moy, R. W. Y.; Price, K. E.; Cummings, M. A.;
(13) Jain, S.; Bates, F. S. Science 2003, 300, 460–464. Roychowdhury, C.; Buffy, J. J.; Olbricht, W. L.; Haaf, M.; McQuade,
(14) Cui, H.; Chen, Z.; Zhong, S.; Wooley, K. L.; Pochan, D. J. D. T. J. Am. Chem. Soc. 2006, 128, 9442–9447.
Science 2007, 317, 647–650. (46) Malmstadt, N.; Nash, M. A.; Purnell, R. F.; Schmidt, J. J. Nano
(15) Wang, X. S.; Guerin, G.; Wang, H.; Wang, Y. S.; Manners, I.; Lett. 2006, 6, 1961–1965.
Winnik, M. A. Science 2007, 317, 644–647. (47) Hautekeer, J.-P.; Varshney, S. K.; Fayt, R.; Jacobs, C.; Jerome,
(16) Forster, S.; Plantenberg, T. Angew. Chem., Int. Ed. 2002, R.; Teyssie, P. Macromolecules 1990, 23, 3893–3898.
41, 688–714. (48) Zhong, X. F.; Varshney, S. K.; Eisenberg, A. Macromolecules
(17) Liu, G. Adv. Polym. Sci. 2008, 220, 29–64. 1992, 25, 7160–7167.
(18) Li, Z.; Kesselman, E.; Talmon, Y.; Hillmyer, M. A.; Lodge, T. P. (49) Yen, B. K. H.; Gunther, A.; Schmidt, M. A.; Jensen, K. F.;
Science 2004, 306, 98–101. Bawendi, M. G. Angew. Chem., Int. Ed. 2005, 44, 5447–5451.

18863 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864

Journal of the American Chemical Society ARTICLE

(50) Oskooei, S. A. K.; Sinton, D. Lab Chip 2010, 10, 1732–1734.

(51) Khan, S. A.; Gunther, A.; Schmidt, M. A.; Jensen, K. F.
Langmuir 2004, 20, 8604–8611.
(52) Gunther, A.; Jensen, K. F. Lab Chip 2006, 6, 1487–1503.
(53) Due to discrepancies between actual and programmed (nominal)
gas flow rates arising from the compressible nature of the gas and the
high gas/liquid interfacial tension, actual gas flow rates were calculated
from from image analysis of the flow and are reported in Supporting
Information (Table S1).
(54) Zaccone, A.; Gentili, D.; Wu, H.; Morbidelli, M. J. Chem. Phys.
2010, 132, 134903.
(55) Zaccone, A.; Wu, H.; Gentili, D.; Morbidelli, M. Phys. Rev. E
2009, 80, 051404.
(56) Melis, S.; Verduyn, M.; Storti, G.; Morbidelli, M.; Baldyga, J.
AIChE J. 1999, 45, 1383–1393.
(57) Zinchenko, A. Z.; Davis, R. H. Phys. Fluids 1995, 7, 2310–2327.
(58) For calculating Pe, D0 was estimated from off-chip dynamic
light scattering data of similar sized PS-b-PAA micelles under the same
temperature and solvent conditions.
(59) Jusufi, A.; Likos, C. N.; Ballauff, M. Colloid Poly. Sci. 2004,
282, 910–917.
(60) Zhang, L. E. A. Macromolecules 1996, 29, 8805–8815.
(61) Grishchuk, A. A.; Estrin, Y. I. Russ. J. Appl. Chem. 2007, 80,
1940–1944.
(62) Assuming that the internal densities and heights of spherical
and nonspherical micelles are the same, the areal percentage of
nonspherical micelles in TEM images should be roughly equivalent to
their weight percentage relative to the entire micelle population.
(63) Dormidontova, E. E. Macromolecules 1999, 32, 7630–7644.
(64) Jones, J. L.; Marques, C. M.; Joanny, J.-F. Macromolecules 1995,
28, 136–142.
(65) This mixing channel geometry run under similar conditions
resulted in complete mixing of the liquid plug within 2 mm of the 10 mm
mixing chamber (ref 41), or 0.5 s at the low flow rate. While this short
channel length (and short duration) is sufficient to mix the reactants in
the liquid plug, it is not sufficient to expose a significant volume of the
liquid to the high shear hot spots in the corners of the liquid plug.
(66) Reiter, G. Phys. Rev. Lett. 1992, 68, 75–78.

18864 dx.doi.org/10.1021/ja2067252 |J. Am. Chem. Soc. 2011, 133, 18853–18864