Soft Matter

View Article Online

PAPER View Journal | View Issue

Equation of state of colloidal membranes

Cite this: Soft Matter, 2019,
Andrew J. Balchunas,a Rafael A. Cabanas,a Mark J. Zakhary,a Thomas Gibaud, b

Seth Fraden, a Prerna Sharma, c Michael F. Hagan a and Zvonimir Dogic *ad
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

15, 6791

In the presence of a non-adsorbing polymer, monodisperse rod-like colloids assemble into one-rod-length
thick liquid-like monolayers, called colloidal membranes. The density of the rods within a colloidal membrane
is determined by a balance between the osmotic pressure exerted by the enveloping polymer suspension
and the repulsion between the colloidal rods. We developed a microfluidic device for continuously observing
an isolated membrane while dynamically controlling the osmotic pressure of the polymer suspension. Using
this technology we measured the membrane rod density over a range of osmotic pressures than is wider
Received 25th May 2019, that what is accessible in equilibrium samples. With increasing density we observed a first-order phase
Accepted 26th July 2019 transition, in which the in-plane membrane order transforms from a 2D fluid into a 2D solid. In the limit of
DOI: 10.1039/c9sm01054h low osmotic pressures, we measured the rate at which individual rods evaporate from the membrane. The
developed microfluidic technique could have wide applicability for in situ investigation of various soft
rsc.li/soft-matter-journal materials and how their properties depend on the solvent composition.

Introduction viruses assembled into hexagonally shaped membranes.6 More

recently, X-ray scattering analysis confirmed crystalline order
Colloidal membranes are one-rod-length thick monolayers of for these conditions.7 Solid membranes have also been
aligned rods that assemble in a presence of depleting polymer.1–4 observed in a mixture of virus and poly-ethylene-glycol, whose
They represent a robust pathway for assembly of self-limited osmotic pressure is temperature dependent.8 This feature
structures that does not rely on the chemical heterogeneity of made it possible to induce nucleation and crystallization
the amphiphilic building blocks, but rather on their geometry. in situ, which was accompanied by large 3D out-of-plane
Numerous properties of colloidal membranes, including their out- membrane deformations. Another work examined the liquid-
of-plane bending rigidity, are determined by the in-plane density of to-crystal transition with X-ray scattering techniques.9 Our
the rodlike constituents.1,5 In turn, this density is determined by measurements reveal the magnitude of the discontinuous
the osmotic pressure exerted by the enveloping depletant that is volume change at the transition, and allow us to fit the entire
balanced by the electrostatic repulsions of the charged rod-like equation of state to a theoretical model. These results have
viruses. We describe a microfluidic platform that allows us to implications for understanding colloidal membranes. For example,
continuously change the depletant concentration, while simulta- the bending modulus of a colloidal membrane is characterized by
neously measuring the membrane area using optical microscopy. the local compression or expansion of molecules, which increases
From such data we reconstruct the colloidal membrane equation away from the stress-free neutral surface.10 Thus, the out-of-plane
of state, which relates the osmotic pressure exerted on the membrane deformations are intimately coupled to its in-plane
membrane to the effective in-plane rod density. With increasing compressibility modulus. The equation of state yields the lateral
osmotic pressure, the equation of state exhibits a discontinuity due compressibility, which in turn provides an estimate of the curvature
to a 2D liquid-to-solid phase transition of the constituent rods. modulus of the colloidal membranes.
Colloidal membranes with in-plane crystalline order were
studied previously. Early studies have demonstrated that in the
presence of small molecular weight depletant, filamentous Materials and methods
Microfluidics device for in situ buffer exchange
a
Department of Physics, Brandeis University, Waltham, MA 02454, USA Colloidal membranes are fragile structures held together by
b
Unversité de Lyon, Ens de Lyon, Université Claude Bernard, CNRS,
weak osmotic pressures. Even the slightest flows, or the move-
Laboratoire de Physique, F-69342 Lyon, USA
c
Department of Physics, Indian Institute of Science, Bangalore 560012, India
ment of an AFM tip close to the surface can fragment a
d
Department of Physics, University of California at Santa Barbara, Santa Barbara, membrane. To overcome these challenges we designed a micro-
CA 93106, USA. E-mail: zdogic@ucsb.edu fluidic device that exchanges the buffer with minimal flow

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 6791--6802 | 6791
 View Article Online

Paper Soft Matter

distortions, while allowing for continuous observation of the reservoir were not directly connected. The water reservoir was
assemblage with high numerical aperture optics. Technologies critical because of the long duration of the experiments and the
have been developed that equip microfluidic devices with small sample volumes. The samples were visualized with high
dialysis membranes that enable rapid spatial and temporal numerical aperture objective. The device thickness required a
buffer exchange.11,12 However, our experiments impose an long working distance condenser for brightfield and differential
additional constraint because of the need to exchange large interference contrast (DIC) microscopy. High resolution images of
molecular weight depletant polymers that do not pass through the membranes were obtained using fluorescence.
conventional dialysis membranes. Directly flowing fluid into the chamber resulted in air
Our microfluidic device had two PDMS layers bound bubbles trapped in the notches. Eliminating air from the device
together. The first layer contained the long channel with was accomplished through dead-end filling the device by closing
stepped notches on both sides, while a second layer contained the outlet valve and then using a syringe pump to load the
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

pressurised channels that act as valves that open and close the chamber. As PDMS is highly permeable to air, the chamber was
flow channel (Fig. 1).13,14 The notches reduced the laminar flow evacuated by the air diffusing through the closed valves. Once
velocity. Due to the continuity of the flow streams, larger loaded with a rod-polymer mixture, the formation of membranes
channel cross-sections reduced flow velocities. We performed with diameter of tens of microns required about 24 hours.
experiments in notches with one or two steps, but membranes Glass coverslips (GoldSeal, FisherScientific) were coated
stored in a notch with no steps were typically destroyed by the with the acrylamide polymer brush to prevent virus adsorption.15
laminar flow through the main channel. A critical element was Plasma treatment usually used for bonding PDMS to glass
the inclusion of the second layer that contained two on-chip destroyed the polymer-brush. We tried, but failed, to grow the
valves. When closed, these effectively blocked minute flows polymer brush after bonding the PDMS to glass. Therefore, we
during the many-hour long interval that is necessary for the clamped together the PDMS chip to the acrylamide treated glass
membranes to grow to a sufficient diameter. At a flow rate of coverslip using an aluminum frame (Fig. 1a). Every component
10 mL per hour, a complete buffer exchange took approximately in the device can be washed and reused. To avoid bending
ten minutes, including the time for the depletant to diffuse deformations of the thin PDMS layers when clamping the device,
from the main channel into the notches (Fig. 2). a slab of rigid Mylar plastic and a shaped shim plate of COC
The experiment required maintaining the same buffer com- plastic layer were used to apply pressure directly over the chip
position over many hours. To accomplish this, the device was elements forming the seal. To ensure uniform compression of
designed with a cavity that encircled the channel and notches the layers, and to provide compliance while sealing, a soft rubber
and acted as a aqueous reservoir that is included in the first O-ring was placed on top of the stack in contact with the
layer. This PDMS layer was sandwiched between a microscope aluminum clamp. The aluminum clamp was tightened using
glass slide and a plastic COC slab with low water permeability, four screws. Enough pressure had to be applied to avoid leakages
thereby minimizing the evaporation of water from the sample from the main channel or the reservoir, but too much pressure
chamber to the outside (Fig. 1). A hydrostatic pump was created yielded collapsed channels.
by closing the outlet and connecting the inlet to a solvent Manufacture of the microfluidic device required two photo-
filled tube that was positioned 1 m above the sample. This resist masters that were used to mold the PDMS device. The
replenished the evaporated solvent due to PDMS permeability. masters were fabricated using conventional soft photolithography
The main channel that contained the sample and the water techniques.16 The flow channel and the reservoir were produced

Fig. 1 The microfluidic device for buffer exchange. (a) Image of a fully sealed and assembled microfluidic device. The PDMS chip is clamped to a glass
coverslip, using two pieces of aluminium and four screws, to ensure no leakage between the two layers. (b) A schematic cross section of the microfluidic
device with individually labelled layers. A rubber O-ring and plastic spacer distribute the clamping pressure across the PDMS chip evenly. (c) Flow is from
right to left. A top-down view of the device design debossed on the PDMS chip. The design has an inlet and outlet (red), a flow channel with storage
notches (black), and two Quake valves (blue) that control the flow through the channel. Small circular posts, placed at the entrance of each notch,
combined with a step-shaped design, reduce laminar flow velocity in the notches. (d) Flow is from right to left. A brightfield image of a PDMS device filled
with depletant and buffer. The field of view is at the centre of the flow channel. (e and f) The valve that controls the flow within the channel, shown in
open and closed configuration. Scale bars, 1 cm (a), 1.5 mm (c), and 250 mm (d–f).

6792 | Soft Matter, 2019, 15, 6791--6802 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Soft Matter Paper

Fig. 2 Time delay associated with buffer exchange. (a) The microfluidic exchange device is viewed with fluorescence microscopy. The fluorescent
depletant and buffer mixture is flowed in from the right side of the page. Both valves are in the open configuration. The time after flow was started is
presented on each frame as min:sec. Measurements are performed in a small area of a one-step notch (shaded in blue in the last time frame). (b) Pixel
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

intensity measures the concentration of the new mixture in the device, relative to the original sample concentration. The complete buffer exchange
occurs after the intensity plateaus. The mean pixel intensity, with background subtraction, was measured in the shaded blue region as a function of time.
Scale bar, 250 mm.

from one master, with photoresist features of two different purified using standard methods.20 The viruses were prepared at
heights with one using positive photoresist and the other negative isotropic–nematic phase coexistence, and the experiments were
photoresist. The inlet and the outlet valves were deposited in a performed by only using viruses from the isotropic fraction.
positive photoresist to achieve a B10 mm height. The main Shorter rods preferentially partition into the isotropic phase,
channel with the notches and the reservoir, were deposited using hence this procedure reduced the number of long, end-to-end
negative photoresist with B50 mm height. A separate wafer fused oligomers that destabilize colloidal membranes.1 A deplet-
encoded the control layer for the two valves. Once molded into ing agent, Dextran MW 500 000 (Sigma-Aldrich, D5251, Lot
the PDMS, these pressure driven control valves were positioned #023K0675), was mixed with viruses at conditions known to
above the inlet and outlet valves. The PDMS device was fabricated promote assembly of colloidal membranes. All samples were
by spin coating a thin layer of PDMS over the wafer that contained made in 100 mM NaCl, 20 mM Tris–HCl, pH 8.0 buffer, and had
the inlet, outlet, main channel and reservoir. A several mm thick a final virus concentration of 2.5 mg mL1. Due to the difference
layer of PDMS was poured over the wafer that contains the in persistence length of fd-wt and fd-Y21M, the initial amount of
features of the valves. Unlike the original Quake formulation,13 depletant was different for each virus type:19 fd-wt and fd-Y21M
we used 1 : 10 ratio of PDMS components for both layers. Once membranes were assembled at depletant concentrations of
cured, both layers were bonded using plasma, and inlet and outlet 42 mg mL1 and 37 mg mL1, respectively. The mixture was
fill-holes were punched into the PDMS. This way, the inlet, outlet, transferred into tubing (PTFE #30 AWG, Cole-Parmer) using a
main channel with notches and reservoir were exposed and syringe (5 mL, Hamilton Gastight), and a syringe pump (PHD
clamped directly on top of the microscope slide, while the 22/2000, Harvard Apparatus) was used to dead-end fill the PDMS
valves that open and close the inlet and outlet channels where device. The sample was always introduced at constant flow-
situated directly above and oriented perpendicular across those rate, so the pressure in the device channels is unknown.
channels (Fig. 1). Fluorescent membranes were assembled with viruses labelled
with an amine reactive fluorescent dye (DyLight-550 NHS Ester,
Optical microscopy, data acquisition and analysis FisherScientific).21 There were about 20 dye molecules per virus
Experiments were performed using an inverted optical micro- as determined by the relative intensity of the absorbance at the
scope (Nikon Eclipse Ti) equipped with a long working distance wavelength related to the virus (269 nm) and the fluorescent dye
(LWD) condenser, and a 100 oil immersion objective (PlanFluor, (550 nm). Samples used to characterize the dynamics of the
NA 1.3). A optical tweezer, created by focusing a 1064 nm laser constituent rods were doped with labelled viruses at a ratio of
(Compass 1064, Coherent), and steered with a pair of acoustic- 1 to 50 000 of labelled to unlabeled particles. Fluorescence images
optic modulators was used to position membranes in the micro- revealed the dynamics of individual rods within the membrane.
fluidic notches.17 Images were acquired using a cooled CCD
camera (Clara, Andor) controlled by Micro-Manager. The colloidal Image analysis
membranes were visualized either with DIC or fluorescence micro- The membrane area was measured using two methods. First,
scopy. For fluorescence images, the microscope was equipped with we determined the area of a fluorescently labelled membrane
a mercury light source (X-Cite 120Q) and a single band dichroic with ImageJ, by setting an intensity threshold at two standard
filter (LED-TRITC-A-NTE-ZERO, SemRock). deviations above the mean pixel intensity. Second, we used DIC
microscopy to visualize the membrane’s edge, and from there
Sample preparation we measured the membrane area with the circle tool in ImageJ.
Bacteriophage fd-wt (wild-type) and fd-Y21M have almost the same Both measurements were consistent with each other. Due to its
contour length of B900 nm.18 The persistence length of fd-wt is 2. monolayer structure, the membrane area is directly propor-
8 mm while that of fd-Y21M is 9.9 mm.19 Both bacteriophages were tional to its volume. To convert the measured area into density,

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 6791--6802 | 6793
 View Article Online

Paper Soft Matter

we analysed a few equilibrium membrane samples with small

angle X-ray (SAXS) measurements, and extracted the particle
separation. Interactive data language (IDL, ITT VIS) software
was used to track individually labelled viruses, and from there
measure the mean square displacement for isolated rods
diffusing within a membrane.

SAXS experiments
Small angle X-ray scattering (SAXS) experiments on individual
membranes were performed at the ID02 beamline of the
European Synchrotron Radiation Facility in Grenoble, France.22 Fig. 3 Three views of colloidal membranes. (a) A top-down view of a 2D
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

The samples were held in the cover slip chamber with membranes colloidal membrane imaged using DIC microscopy. (b) A schematic of a
colloidal membrane. There is complete phase separation between the
lying flat on the coverslip. The chamber was set perpendicular to
self-assembled rod-shaped viruses (red) and depleting polymer, which
the X-ray device, so that the membranes were in a face-on maximizes the system entropy (polymer is not shown). Scale bar, 5 mm.
configuration with respect to the incoming beam. The sample
was scanned with a 20  20 mm2 X-ray beam for individual
colloidal membranes, and signal was recorded with a q range the mesoscale disks grew into mature membranes through
between 0.04 nm1 to 2 nm1. The intensity measured in the lateral coalescence.23 After reaching a certain size, the dense
supernatant was used as the background signal. The plotted membranes sedimented on to the coverslide, ensuring that
intensities were radial averages of the difference between the they remained in the image plane for the duration of the
scattered intensity from the membrane and the background experiment (Fig. 3). We formed colloidal membranes within a
intensity. microfluidic device. Subsequently, using an optical tweezer,
X-Ray experiments yield a scattered intensity I(q) with a we placed a few isolated circular membranes within a stepped
correlation peak positioned at qpeak, which is related to average notch on a side of the main channel. The membranes were
lateral filament separation (Fig. 4b). To estimate the average continuously visualized with either fluorescence, brightfield, or
distance between the filaments we need to divide the measured DIC microscopy. For conditions where equilibrium membranes
intensity by the form factor F(q) which would yield the structure formed, the mature membranes had a constant area as long as
factor: S(q) B I(q)/F(q). The form factor for our experiment the buffer composition remained the same, indicating that the
would consist of a dilute suspension of rods that are all aligned evaporation of the virus into the isotropic suspension was
along the along the direction of the incident X-ray beam. negligible on the experimental time scales. Consequently, any
In practice, this is not doable as dilute rod suspensions are change in the membrane area in response to changing osmotic
necessarily isotropic (Fig. 4a). To make progress we theoretically stress, or equivalently depletant concentration, was directly
estimate the form factor and from there calculate the structure related to the change in the virus density, within an unknown
factor of colloidal membranes. Note that the division by the form scaling factor.
factor shifts the peak location by at most 3% (Fig. 4b). From the We performed a sequence of buffer exchanges. After each
measured structure factor we estimate the average filament exchange, we waited B60 minutes to allow the depletant to
separation d = 2p/qpeak. Assuming local hexagonal packing yields diffuse into the stepped notch and the sample to equilibrate,
the density of rods within fd-wt membranes. before measuring the membrane area. The area significantly
To normalize the fd-Y21M equation of state, we estimated changed with depletant concentration. For example, increasing
the densities of these membranes using DIC microscopy. The the Dextran concentration from 33.5 mg mL1 to 100 mg mL1
DIC signal along the shear axis is proportional to the index of reduced the area by 42% (Fig. 5). We measured the area of the
refraction difference between the membrane and the buffer. In same membrane for B20 different dextran compositions.
turn, the membrane index of refraction is linearly proportional In order to average data from different membranes all measure-
to the virus density. The edge intensity along the shear axis ment sweeps had at least one data point with a common
and the membrane’s rod density should be linearly related. Dextran concentration, which was used to rescaling the area.
Comparing this data with the DIC edge intensity of a fd-wt The above-described procedure yielded the membrane area
membrane of a known density, yielded the absolute density of as a function of the applied osmotic pressure. Converting this
fd-Y21M membranes. Repeating this analysis on fd-wt membranes data into an equation of state requires the scaling factor that
with known densities from X-ray diffraction measurements relates area to virus density. Therefore, we have measured the
revealed that this method has B10% accuracy. virus density within a membrane by using small angle X-ray
scattering, which yielded an isotropic ring whose peak location
Experimental results can be used to calculate the average lateral filament separation
After mixing the virus and polymer suspension, one-rod length (Fig. 4). The scattering experiments were repeated at a few
thick mesoscopic disk-like structures nucleated in the bulk dextran concentrations for which equilibrium membranes
suspension. Within a few minutes, viruses completely phase formed, and the densities found in the X-ray were consistent
separated from the isotropic polymer suspension. Thereafter, with those found in the microfluidic measurements.

6794 | Soft Matter, 2019, 15, 6791--6802 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Soft Matter Paper

Fig. 5 Membrane area decreases with increasing osmotic pressure. A series

of images of a particular membrane, composed of fluorescently labelled
fd-Y21M rods, shown at four different dextran concentrations. Numerical
labels indicate dextran concentration in mg mL1. Scale bar, 5 mm.
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

to solid-like membranes. The transition is reversible, with no

measurable hysteresis.
We also measured the equation of state for fd-Y21M mem-
branes, a virus with a point mutation in the major coat protein
(Fig. 6b). This subtle structural change yields filaments with
essentially the same contour length, but a larger persistence
length of 9.9 mm.18,19 Both fd-wt and fd-Y21M have similar
equations of state, although the dextran concentration required
to induce a phase transition into a 2D crystal is different for the
two virus types. Fd-wt membranes crystallized at B55 mg mL1
while fd-Y21M crystalized at B42.5 mg mL1. We ascribe
the difference in the transition pressure to the difference in
Fig. 4 Small angle X-ray scattering on colloidal membranes. (a) Form persistence length of the viruses.
factor of an isotropic dispersion of fd-virus at 0.5 mg mL1 (circle). We investigated the upper and lower bounds over which
Corresponding theoretical form factor F of the fd-virus all aligned along
the colloidal membranes remain metastable. At sufficiently
the X-ray beam. (b) Intensity scattered by a colloidal membrane (bottom
curves) and structure factor S = I/F (top curves) with F being the theoretical high osmotic pressures we found that colloidal membranes
form factor. The position of the peak is indicated by the vertical dash line. solidified and fractured. Notably the onset of fracture depended
Going from I to S and increasing dextran concentration c shift qpeak from on virus type: fd-wt fractured at much lower concentrations
0.517 to 0.530, 0.530 to 0.540 and 0.545 to 0.553 nm1. compared to fd-Y21M. At lower dextran concentrations we
found two different pathways by which membranes became
unstable. In one pathway, we observed direct nucleation of 3D
Equilibrium colloidal membranes of fd-wt formed over a nematic tactoids. The nucleation event induced formation of
fairly narrow range of dextran concentrations (from 40 mg mL1 large-scale flows that quickly transported all the rods from a
to 55 mg mL1). The microfluidic device allowed for measure- metastable 2D membrane into a more stable 3D tactoid (Fig. 8).
ments over a significantly larger range of osmotic pressures Intriguingly the tactoid nucleation events only occurred at the
(Fig. 6a). This was possible because 2D colloidal membranes are membrane’s edge, where the boundary conditions induce tilt
highly metastable structures that cannot readily melt into 3D and the associated formation of a thin edge-bound layer of a
nematic tactoids at lower depletant concentrations, or freeze into quasi 1D-nematic phase.23,24 The alternate pathway is direct
3D smectic liquid crystals at high concentrations. The equation of evaporation of rods into the background isotropic suspension
state exhibited a pronounced discontinuity at 55 mg mL1 dextran at low enough osmotic pressures (Fig. 9).
concentrations, where the area changed by 8%. This is suggestive To investigate rod evaporation, we formed fd-Y21M membranes
of a first order phase transition, due to two-dimensional freezing in the stable regime at a dextran concentration of 40 mg mL1.
of the in-plane virus order. At osmotic pressures below the Subsequently, we performed a buffer exchange to a dextran
transition, colloidal membranes exhibited significant edge concentration to 25 mg mL1, a regime where the equation of
fluctuations due to finite edge tension.24 Increasing the dextran state could not be measured since the membrane area changes on
concentration suppressed these fluctuations, and above the experimental time scales. The buffer exchange took B10 minutes,
transition they entirely ceased. The arrest of edge fluctuations while the membrane area decreased on a longer time scale of tens
was accompanied by a discontinuous change in the membrane of minutes (Fig. 9). In principle, virus particles can evaporate either
area. Doping the membrane with fluorescent rods revealed the from the edge-bound 1D nematic, or directly from the 2D
dynamics of constituent rods in both phases. At lower osmotic membrane interior. The former case would yield the area rate of
pressures viruses exhibited diffusive motion, but above the change proportional to the membrane circumference, or equiva-
discontinuity their motion ceased (Fig. 7). This confirms our lently the square root of the area. For the latter case, the evapora-
hypothesis that with increasing osmotic pressure colloidal tion rate would be proportional to the area. The experimentally
membranes undergo a freezing transition from a liquid-like measured curve was well described by the exponential curve, which

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 6791--6802 | 6795
 View Article Online

Paper Soft Matter

Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

Fig. 6 Equation of state of fd-wt and fd-Y21M colloidal membranes. (a) Rod density as a function of dextran concentration measured for fd-wt
membranes. Inset: Discontinuity in the equation of state at B54 mg mL1 is consistent with a first order freezing and phase transition between a 2D
liquid-like and 2D solid-like membrane. Diamonds indicate three samples where the in-plane rod density was determined with small angle X-ray
scattering measurement. Red bars indicate regions where colloidal membranes are equilibrium structures. (b) Equation of state for fd-Y21M membranes.
The membrane freezes at smaller osmotic stresses. The orange lines are guides to the eye. Error bars are the standard deviation of 10 measurements of
membrane area.

there is no depletant inside the membrane. Therefore, mechanical

equilibrium requires that the applied osmotic pressure, Papp, be
balanced by the pressure of the rods within colloidal membrane,
Pint. We estimate the osmotic pressure arising from inter-rod
electrostatic interactions and the suppression of rod conforma-
tional fluctuations as the rod density increases using a previous
theoretical model.25 In brief, we calculate the lateral extent of
conformational fluctuations, d, of a filament within a confining
electrostatic field generated by nearest neighbour rods arranged on
a hexagonal array. The characteristic length along the rod contour
Fig. 7 Dynamics of viruses in liquid and solid membranes. (a) A top-down
view of a fd-wt membrane (red) taken with DIC microscopy is overlaid with of such undulations is given by the deflection length:
a fluorescence image that shows isolated viruses (cyan). This technique
was used to visualize individual rods moving in the bulk. (b) The mean
ld D d2/3lp1/3 (1)
square displacement of fluorescently labelled rods was measured as a
function of time for both a 2D liquid-like (green) and a 2D solid-like
with lp the filament persistence length.26–28 Each deflection
(purple) membrane. Rods in the liquid-like membrane exhibit diffusive gives rise to a free energy cost that scales with kBT, so the free
behaviour with a diffusion coefficient of 0.03 mm s1, while those in the energy per rod arising from conformational fluctuations
frozen membrane have no measurable motion. Scale bar, 5 mm. is proportional to the number of deflections over the total
filament contour length:
indicates that virus mainly evaporated throughout the entire
Fconf = ckBT/ld (2)
membrane interior. The unbinding rate for the viruses under these
conditions is on average 4.1 min1. with the constant given by c = 2 2/3 25
.
We assume that the deflection length is large compared to
Theoretical model of the equation of state the Debye length ld c k1 so that the filaments can be locally
We develop a phenomenological model that explains some treated as rigid rods when calculating the electrostatics. In this
aspects of the measured equation of state. We assume that limit, the far field of the electrostatic potential f from a rod

Fig. 8 A 2D colloidal membrane melts into a 3D nematic tactoids. A time lapse showing how a metastable 2D colloidal membrane melts into a stable 3D
tactoid (blue dashed outline) at low osmotic pressure. The tactoid nucleates from the membrane’s twisted edge, where nematic order is already very
high. The tactoids exhibit a surface frozen smectic monolayer. Scale bar, 10 mm.

6796 | Soft Matter, 2019, 15, 6791--6802 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Soft Matter Paper

semiflexible regime because it assumes a persistence length

that is small or on the order of the inter-rod spacing, as has
been previously noted.25 A more recent calculation for the effect
of fluctuations on inter-rod interactions obtains a decreasing
dependence of osmotic pressure with persistence length.33
However, we do not observe the predicted doubling or quad-
rupling of the apparent decay length of the interactions due to
fluctuations for the range of inter-rod separations studied in
our experiments, although this could become relevant for larger
inter-rod separations.
Fig. 9 Evaporation of colloidal membranes at low osmotic pressures. To proceed further, we need to solve for the effective linear
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

(a) Membrane area as a function of time was measured for different

charge density xeff. Due to the finite width of the fd-virus, the
membranes. Two curves were normalized such that evaporation begins
at area equal to one, and experiments were performed at 25 mg mL1 estimate for an infinitely thin cylinder that counterions will
dextran concentration. Each curve represents a unique membrane. renormalize the charge density to one charge per Bjerrum length
All curves are fit well by exponential decay functions and exhibit an average is inapplicable.34 Instead, we note that the Debye–Hückel approxi-
of rate off koff rate of 4.1 min1. (b) A sequence of images showing how the mation accurately describes the far-field form of the electrostatic
membrane area shrinks with time due to evaporation of virus particles.
potential, but over-predicts the potential in the near field.27,35
Scale bar, 5 mm.
Therefore, we find the effective charge density for which the
far-field potential is correct, as has been described previously.
(outside of its double layer) is described by the Debye–Hückel We used an approximate analytical solution to the nonlinear
approximation, which for a cylinder is given by: Poisson–Boltzmann equation around a cylinder, which matches
s a near-field solution to the Debye Hückel far-field.36 Equating the
cðkrÞ  2xeff K0 ðkrÞ; for r  þ k1 ; (3) far-field result to eqn (3) yields the effective charge density, as a
2
function of the bare charge density n0 and k.
where c = ef/kBT is the non-dimensional electrostatic potential, To capture the experimental observation that the decay
e is the elementary charge, K0 is the zero-order modified Bessel length of the interactions is shorter than the bulk Debye length,
function of the second kind, and xeff = lBneff is the dimension- it was necessary to account for the presence of excess counter-
less linear charge density, with lB E 0.71 nm the Bjerrum ions within the colloidal membrane. We related the local Debye
length and neff the effective filament linear charge density.29 length k1 to the bulk value kD1 and the membrane charge
The effective charge density can be calculated from the non- density using the cylindrical cell model.31,37 In this approach,
linear Poisson–Boltzmann equation as described below. the hexagonal Wigner–Seitz cell associated with each virus is
The electrostatic free energy depends sensitively on the scale represented by a cylinder with the same volume. The radius of
of lateral undulations d, since these bring neighbouring rods the effective cylindrical Wigner–Seitz cell is:
closer together, thus enhancing the electrostatic energy. rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi
pffiffiffi.
Performing a variational calculation, in which d is determined RS ¼ drod 3 2p; (6)
by minimizing the total free energy Ftot arising from conforma-
tional fluctuations and electrostatics, results in the following and the local Debye length is calculated as described
expression for d:25 previously,31
1 2 2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
d 8=3 e2k d 2clB drod 1=2 ekdrod k ¼ kD cosh cS ; (7)
¼ : (4)
1 1=2 2 1=3 3=2
1 þ k2 d 2 drod 1 9ð2pÞ xeff lp k and
2
2xeff ðI0 ðkRS ÞK1 ðkRS Þ þ I1 ðkRS ÞK0 ðkRS ÞÞ
Assuming hexagonal ordering of the rods, the osmotic pressure tanh cS ¼ (8)
I1 ðks=2Þ
@Ftot I1 ðkRS Þ  K1 ðkRS Þ
is given by: Pint ¼  1=2 ,30 which yields: K1 ðks=2Þ
3 drod @drod
2ckB T where cS is the dimensionless electrostatic potential at the
Pint ¼ : (5) surface of the cell. Since xeff and k are interrelated, it is necessary
32=3 kdrod d 8=3 lp1=3
to solve for them self-consistently as functions of drod.
A subsequent work extended the Odijk’s calculation and We set the fd diameter as s = 6.6 nm, and use the estimate of
relaxed some approximations; however, we were not able to v0 = 7 e nm1 for fd.38 For the bulk Debye length kD1 = 1 nm,
match the scaling of the experimental applied osmotic pressure we obtain an effective dimensionless linear charge density of
using this model.31 Moreover, an earlier calculation within a xeff = 36.15. For comparison, using the Debye Huckel approxi-
hexagonal layer of charged filaments predicts that the osmotic mation without the near-field correction would result in an
pressure increases with persistence length,32 in contrast to lB
effective charge density of xDH ¼  53:56: The
our experiments which exhibit a decrease in pressure with bkðs=2ÞK1 ðks=2Þ
persistence length. This calculation is inappropriate for the Debye screening length and effective filament charge depend

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 6791--6802 | 6797
 View Article Online

Paper Soft Matter

Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

Fig. 10 The electrostatic screening length and effective filament charge Fig. 11 The theoretical predictions for the osmotic pressure as a function
within colloidal membranes. The inverse Debye length, k (black line), of inter-rod separation are shown for persistence lengths corresponding
and the corresponding dimensionless effective linear charge density to fd-wt (lp = 2 mm, blue line) and fd-Y121M (lp = 10 mm, orange line),
xeff (dashed red line) are plotted as a function of inter-rod separation. overlaid on the experimental data. The theoretical results are obtained
The effective charge density is normalized by its bulk value, xeff = 36.15 for from eqn (4)–(9). The dashed black line shows the result in the rigid rod
kD = 1 nm1. limit (lp = N).

on the filament separation (Fig. 10). The local Debye length varies
by 40% over the experimental range of inter-rod distances.
To compare the experimental data to the theoretical model
we converted the dextran concentration into the osmotic pres-
sure using a modification of the previously published empirical
relationship:

P(c) = A1(mw0/mw)c + A2c2 + A3c3, (9)

where c is the dextran weight fraction, A1 = 0.0655 atm cm3 gm1,

A2 = 10.38 atm cm6 gm2, A3 = 75.3 atm cm9 gm3, P is the
osmotic pressure in atmospheres, mw = 6.7  105 g mol1
is the dextran molecular weight in our experiments, and mw0 =
3.7  105 g mol1 the dextran molecular weight used for the
published measurement.39 The term mw0/mw corrects the van’t Fig. 12 Osmotic pressure of crystalline membranes scales exponentially
with filament separation. The experimentally measured equation of state
Hoff coefficient for the molecular weight difference. Note that
for fd-wt and fd-Y121M is plotted as a function of the inter-rod separation.
the osmotic pressure is relatively insensitive to molecular The log scale Y-axis illustrates the exponential decay. The solid black line is
weight at experimentally relevant concentrations.39–41 Impor- a best fit to the data within the crystalline regime, yielding a decay length of
tantly, this relationship shows that the osmotic pressure is non- keff1 = 1.04 nm.
ideal, exceeding the van’t Hoff ideal gas limit by more than an
order of magnitude.
The above described model assumes in-plane hexagonal the observation that the measured equation of state is nearly
order. This is strictly applicable only in the crystalline phase, insensitive to persistence length (for lp Z L) in the limit of small
although there will still be local hexagonal order in the liquid inter-rod spacing, or equivalently high osmotic pressures. For
phase. Plotting the experimental data for fd-wt and fd-Y12M comparison, the theoretical result in the rigid rod limit is also
against the theoretical prediction for the applied pressure shown. In this regime, filament undulations are suppressed to
reveals qualitative agreement in several respects (Fig. 11). First, scales that are smaller than the Debye length.
the theory captures the apparent exponential decay, with decay The theory over-estimates the osmotic pressure even in the
constant keff1 E 1.07 nm, in agreement with the experimental rigid rod limit by a factor B5. This can be attributed either to
decay length, keff1 E 1.04 nm (Fig. 12). These are only the use of the Poisson–Boltzmann model, which neglects ion
apparent decay lengths since the local Debye length is a correlations, or neglecting the nonuniformity of the fixed
function of the inter-rod spacing. Measuring pressure rather charges in the viruses, or to the approximate calculation of
than force incurs an extra dependence on inter-rod spacing. the effective linear charge density. For example, the extent of
Applying the same analysis to the force, f = 31/2drodPint(drod), counterion condensation onto charged lamellar stacks increases
yields apparent decay lengths of 1.16 nm and 1.20 nm for with decreasing layer separations, thus leading to charge regulari-
experiment and theory, respectively. Second, the theory reproduces zation that reduces the effective surface charge.42 We chose not to

6798 | Soft Matter, 2019, 15, 6791--6802 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Soft Matter Paper

account for this effect because it is likely small in our case because separation from X-ray scattering, we obtained the rod density
the surface separation is large compared to the Debye length, and by measuring the assemblage area with optical microscopy.
it would require an additional fit parameter – the counterion- Since there is no significant exchange of rods on experimental
virus interaction strength. However, note that since our computer times scales, the membrane area is directly related to the
rod–rod interactions are quadratic in the effective charge, the filament spacing. Second, our technique permits in situ change
discrepancy between theory and experiment would be eliminated of the osmotic stress, which yields more accurate measure-
by reducing the effective charge to about half its value. ments. This feature revealed a discontinuity in the equation of
We performed an alternative calculation in which the viruses are state that is associated with a first other phase transition from a
treated as semiflexible filaments with hard-core interactions, liquid to a solid state. In a similar spirit, a recent study has
where the effect of the electrostatics is accounted for by an utilized the temperature dependence of poly(ethylene glycol) to
‘‘effective diameter’’ that is larger than the intrinsic virus diameter. extract the discontinuous change in the filament spacing
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

However, this approach failed to capture the scaling of the associated with the transition from the cholesteric to line
experimental data with drod or the lack of dependence on persis- hexatic.47 Finally, the ability to tune the osmotic pressure
tence length. in situ enabled assembly of metastable colloidal membranes,
It is possible to estimate the location of the melting transition while compatibility with optical microscopy enabled visualiza-
for an array of semiflexible chains interacting through hard-core tion of the kinetic pathways by which these metastable 2D
interactions.43 Extending this model to include electrostatic inter- colloidal structures transform into 3D materials.
actions is challenging and beyond the scope of the present work.31 Equilibrium colloidal membranes form over a fairly limited
However, we can make a qualitative comparison to the Lindemann range of depletant concentrations. Osmotic pressures beyond
criteria for a 3D crystal of point particles, which states that melting an upper critical value leads to the formation of bulk smectic
occurs when the typical size of particle fluctuations reaches 0.1a, phases, while pressures below a lower critical value lead to
with a the lattice spacing.43 The onset of the melting transition and nematic tactoids.48,49 Our microfluidic technique allowed for
its dependence on persistence length can be roughly captured by assembly and investigations of metastable colloidal membranes
equating the scale of undulations (eqn (4)) to the inter-surface using a two-step process. In a first step we assembled colloidal
separation between rods according to 0.17(drod  s), where the membranes for a range of osmotic pressures where they are
factor 0.17 was fit by eye (Fig. 13). However, this comparison is equilibrium structures. Subsequently, we changed the osmotic
qualitative at best, and the melting transition predicted previously pressure, to values where colloidal membranes do not assemble
required undulations of about five times the Lindemann criteria.43 directly from an isotropic suspension. These experiments revealed
that colloidal membranes are metastable over a wide range
Discussion of osmotic pressures and also demonstrated the fundamental
difference between colloidal membranes and conventional 3D
Osmotic stress techniques have been used to interrogate pro- materials. Formation of either 3D tactoids or bulk smectic phases
perties of diverse soft materials, providing insight into the from a 2D colloidal membranes requires a rearrangement in
molecular interactions that govern behaviors of both lipid which rods collectively escape into the third dimension. Such
bilayers and biopolymer suspensions.44–46 Our method has a transformations have large nucleation barriers. At sufficient low
few differences with the standard implementation of the osmotic pressures, a membrane can either directly evaporate into an
stress techniques. First, instead of extracting the average filament isotropic phase or melt into a 3D nematic tactoid. Tactoid
nucleation always takes place at the membrane’s edge. This can
be rationalized as the boundary conditions enforce uniform twist
of the membrane’s edge that locally melts a smectic monolayer
into a nematic.24,50 Similar pathways were observed in mixtures of
viruses and thermosensitive polymer.51,52
Intermolecular interactions. Confined liquid crystalline
semi-flexible filaments have more restricted degrees of freedom
than filaments in a disordered isotropic suspension. Conse-
quently, there is an entropic penalty associated with the formation
of orientationally ordered phases. This penalty is smaller for more
rigid filaments; thus, the stiffer fd-Y21M filaments form mem-
branes at lower osmotic stress, when compared to the more
flexible fd-wt. Our results suggest that theories of colloidal mem-
branes phase behavior which do not account for semi-flexibility
will under-predict the osmotic pressure required to stabilize
Fig. 13 The scale of filament undulations, d, calculated from eqn (4) is
membranes. This is consistent with previous findings that
shown for fd-wt (blue line) and fd-Y121M (orange line) as a function of
inter-rod separation. The straight line is an effective Lindemann criteria,
increasing filament flexibility supressed the formation of both
d = 0.17(drod  s), which estimates the inter-rod separation at which nematic and smectic liquid crystals.19,53–58 Intriguingly, the fila-
the membrane melts for each case. ment flexibility does not influence the equation of state at higher

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 6791--6802 | 6799
 View Article Online

Paper Soft Matter

osmotic pressures or equivalently smaller filaments separations. bending modes. Recently, it has been shown that colloidal
Recent experiments using classical osmotic stress techniques membranes exhibit significant out-of-plane edge-bound modes
examined the interactions of both fd-wt and fd-Y21M virus, and that create saddle-splay deformations. These fluctuations are
reached the same conclusion.38 While our experiments studied a driven by the low and positive value of the Gaussian modulus of
different range of osmotic stresses, our data is consistent with colloidal membranes, and decay as one moves away from the
these results. edge.62 Both the edge fluctuations and a 1D cross-section of the
A Poisson–Boltzmann calculation that accounts for the height fluctuations scale as B1/q3. The only way to rigorously
increased density of counter ions within the membrane (com- distinguish between the two fluctuation modes is to analyze
pared to bulk) explains some features of the measured equation how the measured spectrum changes as one moves away from a
of state in the crystalline regime. This result indicates that free edge.
electrostatics are the dominant interaction between rods at The conflicting measurements of curvature modulus
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

surface–surface separation distances in the range 3–6 nm. demonstrate that the continuum properties of colloidal mem-
Consistent with the experimental results, the theory finds that branes are not well understood and further experimentation is
filament conformational fluctuations have a small effect on rod required. In particular, there is a need for methods that
interactions at inter-rod separations in the crystalline regime. use external force to robustly perturb a colloidal membrane
We found that theories which represent electrostatic effects structure in order to measure its bending rigidity. Alternatively,
through hard-core interactions and an ‘‘effective diameter’’ re-examining the height fluctuations of a colloidal membrane
could not describe the experimental measurements. Consistent viewed in the edge-on configuration, and how they depend on
with this observation, previous models of colloidal membranes the distance away from the edge, might provide additional
and rafts that did achieve quantitative agreement with experi- insight into the microscopic origin of previously measured
ments without accounting for electrostatics assumed that the bending fluctuations.
depletant osmotic pressure is given by an ideal equation of
state.4,59 In fact, the empirically measured equation of state for
dextran (eqn (9)) exceeds the ideal formula by an order of
Conflicts of interest
magnitude at experimentally relevant concentrations. There are no conflicts to declare.
Estimating elasticity of colloidal membranes. The measured
equation of state provides insight into the energetic cost of
elastically deforming colloidal membranes. Similar to lipid Acknowledgements
bilayers, the out-of-plane deformations of colloidal membranes
We thank Tom Powers for useful discussions, and Achini
are described by the Helfrich energy that contains two para-
Opathalage for suggestions related to the design of the micro-
meters, Gaussian and mean curvature moduli.60 The Gaussian
fluidic device. We acknowledge support of National Science
modulus of colloidal membranes, k, has been measured using
Foundation through grants (NSF-MRSEC-1420382, NSF-DMR-
two independent methods.61,62 Both methods yielded k, that is
1609742 and NSF-DMR-1905384). We also acknowledge use of
positive and of the order of B100 kBT, suggesting that mem-
Brandeis MRSEC biological synthesis facility, microfluidics
branes can lower their free energy by adopting saddle-splay
fabrication facility and optical microscopy facilities supported
configurations. An early study of colloidal membranes esti-
by NSF-MRSEC-1420382. TG thanks ID02 staff at the ESRF for
mated that k is also a few hundred kBT.1 The equation of state
their help with SAXS experiments.
yields an independent estimate of the curvature modulus, by
1 @A
calculating the lateral membrane compressibility, ka   .
A @P Notes and references
The compressibility is related to the bending modulus, k, by
k = kah2/12, where h is the membrane thickness. The equation 1 E. Barry and Z. Dogic, Entropy driven self-assembly of
of state measurements yield that ka B 230 000 kBT mm2 at nonamphiphilic colloidal membranes, Proc. Natl. Acad.
40 mg mL1 dextran concentration, which leads to a mean Sci. U. S. A., 2010, 107, 10348–10353.
curvature modulus of k B 15 000 kBT, where we assumed that 2 Y. Yang, E. Barry, Z. Dogic and M. F. Hagan, Self-assembly of
membrane thickness is B0.9 mm. 2D membranes from mixtures of hard rods and depleting
The equation of state based estimate of k is very different polymers, Soft Matter, 2012, 8, 707–714.
from the previous measurement.1 In the original experiments, 3 Y. Yang and M. F. Hagan, Theoretical calculation of the
k was estimated by analysing the out-of-plane height fluctua- phase behavior of colloidal membranes, Physical Review E:
tions of membranes that were viewed in the edge-on configu- Statistical, Nonlinear, and Soft Matter Physics, 2011, 84,
ration. However, the distance from the membrane edge was not 051402.
systematically controlled and the height fluctuations were 4 L. Kang, T. Gibaud, Z. Dogic and T. Lubensky, Entropic
visualized by non-confocal techniques that do not effectively forces stabilize diverse emergent structures in colloidal
eliminate out-of-plane signal contributions. It is possible that membranes, Soft Matter, 2016, 12, 386–401.
the out-of-plane fluctuations studied previously were actually 5 L. D. Landau, E. M. Lifshitz, V. Berestetskii and L. Pitaevskii,
associated with the soft edge modes rather than the intrinsic Course of Theoretical Physics: Theory of Elasticity, 1995, pp. 40–44.

6800 | Soft Matter, 2019, 15, 6791--6802 This journal is © The Royal Society of Chemistry 2019
 View Article Online

Soft Matter Paper

6 Z. Dogic and S. Fraden, Development of model colloidal 28 H. Strey, V. Parsegian and R. Podgornik, Equation of state
liquid crystals and the kinetics of the isotropic–smectic for polymer liquid crystals: theory and experiment, Phys.
transition, Philos. Trans. R. Soc., A, 2001, 359, 997–1015. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip. Top.,
7 B. Sung, A. Cotte and E. Grelet, Chirality-controlled crystal- 1999, 59, 999.
lization via screw dislocations, Nat. Commun., 2018, 9, 1405. 29 S. L. Brenner and V. A. Parsegian, A physical method for
8 L. Saikia, et al., Curvature instability of chiral colloidal deriving the electrostatic interaction between rod-like poly-
membranes on crystallization, Nat. Commun., 2017, 8, 1160. ions at all mutual angles, Biophys. J., 1974, 14, 327.
9 T. Gibaud and D. Constantin, Direct Liquid to Crystal 30 R. Podgornik, D. C. Rau and V. A. Parsegian, The action of
Transition in a Quasi-Two-Dimensional Colloidal Membrane, interhelical forces on the organization of DNA double
J. Phys. Chem. Lett., 2018, 9, 4302–4307. helixes: fluctuation-enhanced decay of electrostatic
10 L. D. Landau and E. Lifshitz, Theory of Elasticity, vol. 7, double-layer and hydration forces, Macromolecules, 1989,
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

Course of Theoretical Physics, 1986, 3, 109. 22, 1780–1786.

11 J. S. Paustian, R. N. Azevedo, S.-T. B. Lundin, M. J. Gilkey 31 D. J. Lee, S. Leikin and A. Wynveen, Fluctuations and
and T. M. Squires, Microfluidic microdialysis: Spatio- interactions of semi-flexible polyelectrolytes in columnar
temporal control over solution microenvironments using assemblies, J. Phys.: Condens. Matter, 2010, 22, 72202.
integrated hydrogel membrane microwindows, Phys. Rev. X, 32 R. Podgornik and V. A. Parsegian, Molecular fluctuations in
2013, 3, 041010. the packing of polymeric liquid crystals, Macromolecules,
12 M. Kornreich, M. Heymann, S. Fraden and R. Beck, Cross 1990, 23, 2265–2269.
polarization compatible dialysis chip, Lab Chip, 2014, 14, 33 H. H. Strey, V. A. Parsegian and R. Podgornik, Equation of
3700–3704. state for polymer liquid crystals: Theory and experiment,
13 M. A. Unger, H.-P. Chou, T. Thorsen, A. Scherer and Phys. Rev. E: Stat. Phys., Plasmas, Fluids, Relat. Interdiscip.
S. R. Quake, Monolithic Microfabricated Valves and Pumps Top., 1999, 59, 999–1008.
by Multilayer Soft Lithography, Science, 2000, 288, 113–116. 34 G. S. Manning, Counterion condensation on charged
14 C. Girabawe and S. Fraden, An image-driven drop-on- spheres, cylinders, and planes, J. Phys. Chem. B, 2007, 111,
demand system, Sens. Actuators, B, 2017, 238, 532–539. 8554–8559.
15 A. Lau, A. Prasad and Z. Dogic, Condensation of isolated 35 A. Stroobants, H. Lekkerkerker and T. Odijk, Effect of
semi-flexible filaments driven by depletion interactions, electrostatic interaction on the liquid crystal phase transi-
EPL, 2009, 87, 48006. tion in solutions of rodlike polyelectrolytes, Macromolecules,
16 J. C. McDonald, et al., Fabrication of microfluidic systems in 1986, 19, 2232–2238.
poly(dimethylsiloxane), Electrophoresis, 2000, 21, 27–40. 36 J. R. Philip and R. A. Wooding, Solution of the Poisson–
17 A. Ward, et al., Solid friction between soft filaments, Nat. Boltzmann Equation about a Cylindrical Particle, J. Chem.
Mater., 2015, 14, 583. Phys., 1970, 52, 953–959.
18 E. Pouget, E. Grelet and M. P. Lettinga, Dynamics in the 37 A. G. Cherstvy, A. A. Kornyshev and S. Leikin, Temperature-
smectic phase of stiff viral rods, Physical Review E: Statistical, dependent DNA condensation triggered by rearrange-
Nonlinear, and Soft Matter Physics, 2011, 84, 041704. ment of adsorbed cations, J. Phys. Chem. B, 2002, 106,
19 E. Barry, D. Beller and Z. Dogic, A model liquid crystalline 13362–13369.
system based on rodlike viruses with variable chirality and 38 G. Abramov, et al., Structural effects of single mutations in a
persistence length, Soft Matter, 2009, 5, 2563–2570. filamentous viral capsid across multiple length scales,
20 T. Maniatis, J. Sambrook and E. Fritsch, Molecular Cloning., Biomacromolecules, 2017, 18, 2258–2266.
Cold Spring Harbor Univ. Press, 1989. 39 H. Vink, Precision measurements of osmotic pressure in
21 M. P. Lettinga, E. Barry and Z. Dogic, Self-diffusion of rod- concentrated polymer solutions, Eur. Polym. J., 1971, 7,
like viruses in the nematic phase, EPL, 2005, 71, 692. 1411–1419.
22 T. Narayanan, O. Diat and P. Bösecke, SAXS and USAXS on 40 V. Parsegian, R. Rand, N. Fuller and D. Rau, Methods in
the high brilliance beamline at the ESRF, Nucl. Instrum. enzymology, Elsevier, 1986, vol. 127, pp. 400–416.
Methods Phys. Res., Sect. A, 2001, 467-468, 1005–1009. 41 F. Vérétout, M. Delaye and A. Tardieu, Molecular basis of
23 M. J. Zakhary, et al., Imprintable membranes from incomplete eye lens transparency: osmotic pressure and X-ray analysis
chiral coalescence, Nat. Commun., 2014, 5, 3063. of a-crystallin solutions, J. Mol. Biol., 1989, 205, 713–728.
24 T. Gibaud, et al., Reconfigurable self-assembly through chiral 42 D. Harries, R. Podgornik, V. A. Parsegian, E. Mar-Or and
control of interfacial tension, Nature, 2012, 481, 348–351. D. Andelman, Ion induced lamellar-lamellar phase transition in
25 T. Odijk, Undulation-enhanced electrostatic forces in hexa- charged surfactant systems, J. Chem. Phys., 2006, 124, 224702.
gonal polyelectrolyte gels, Biophys. Chem., 1993, 46, 69–75. 43 J. V. Selinger and R. F. Bruinsma, Hexagonal and nematic
26 R. Granek, From semi-flexible polymers to membranes: phases of chains. II. Phase transitions, Phys. Rev. A: At., Mol.,
anomalous diffusion and reptation, J. Phys. II, 1997, 7, Opt. Phys., 1991, 43, 2922–2931.
1761–1788. 44 V. A. Parsegian, N. Fuller and R. P. Rand, Measured work of
27 T. Odijk, Theory of lyotropic polymer liquid crystals, Macro- deformation and repulsion of lecithin bilayers, Proc. Natl.
molecules, 1986, 19, 2313–2329. Acad. Sci. U. S. A., 1979, 76, 2750–2754.

This journal is © The Royal Society of Chemistry 2019 Soft Matter, 2019, 15, 6791--6802 | 6801
 View Article Online

Paper Soft Matter

45 H. Strey, V. Parsegian and R. Podgornik, Equation of state 53 A. Khokhlov and A. Semenov, Liquid-crystalline ordering in
for DNA liquid crystals: fluctuation enhanced electrostatic the solution of long persistent chains, Phys. A, 1981, 108,
double layer repulsion, Phys. Rev. Lett., 1997, 78, 895. 546–556.
46 D. J. Needleman, et al., Synchrotron X-ray diffraction study 54 A. V. Tkachenko, Nematic-smectic transition of semiflexible
of microtubules buckling and bundling under osmotic chains, Phys. Rev. Lett., 1996, 77, 4218.
stress: a probe of interprotofilament interactions, Phys. 55 J. Tang and S. Fraden, Isotropic-cholesteric phase transition
Rev. Lett., 2004, 93, 198104. in colloidal suspensions of filamentous bacteriophage fd,
47 S. Yasar, R. Podgornik, J. Valle-Orero, M. R. Johnson and V. A. Liq. Cryst., 1995, 19, 459–467.
Parsegian, Continuity of states between the cholesteric - line 56 P. Van Der Schoot, The nematic-smectic transition in suspen-
hexatic transition and the condensation transition in DNA sions of slightly flexible hard rods, J. Phys. II, 1996, 6, 1557–1569.
solutions, Sci. Rep., 2014, 4, 6877. 57 Z. Dogic and S. Fraden, Smectic phase in a colloidal
Published on 29 July 2019. Downloaded by Brown University on 10/2/2024 8:01:22 PM.

48 Z. Dogic, Surface freezing and a two-step pathway of the suspension of semiflexible virus particles, Phys. Rev. Lett.,
isotropic-smectic phase transition in colloidal rods, Phys. 1997, 78, 2417.
Rev. Lett., 2003, 91, 165701. 58 E. Grelet, Hard-rod behavior in dense mesophases of semi-
49 E. Grelet, Hexagonal order in crystalline and columnar flexible and rigid charged viruses, Phys. Rev. X, 2014, 4, 021053.
phases of hard rods, Phys. Rev. Lett., 2008, 100, 168301. 59 L. Kang and T. C. Lubensky, Chiral twist drives raft formation
50 E. Barry, Z. Dogic, R. B. Meyer, R. A. Pelcovits and and organization in membranes composed of rod-like
R. Oldenbourg, Direct measurement of the twist penetration particles, Proc. Natl. Acad. Sci. U. S. A., 2017, 114, E19–E27.
length in a single smectic A layer of colloidal virus particles, 60 W. Helfrich, Elastic properties of lipid bilayers: theory and
J. Phys. Chem. B, 2008, 113, 3910–3913. possible experiments, Z. Naturforsch., C: Biochem., Biophys.,
51 A. Alsayed, Z. Dogic and A. Yodh, Melting of lamellar phases Biol., Virol., 1973, 28, 693–703.
in temperature sensitive colloid-polymer suspensions, Phys. 61 T. Gibaud, et al., Achiral symmetry breaking and positive
Rev. Lett., 2004, 93, 057801. Gaussian modulus lead to scalloped colloidal membranes,
52 A. Modlińska, A. M. Alsayed and T. Gibaud, Condensation Proc. Natl. Acad. Sci. U. S. A., 2017, 201617043.
and dissolution of nematic droplets in dispersions of 62 L. L. Jia, M. J. Zakhary, Z. Dogic, R. A. Pelcovits and T. R.
colloidal rods with thermo–sensitive depletants, Sci. Rep., Powers, Chiral edge fluctuations of colloidal membranes,
2015, 5, 18432. Phys. Rev. E, 2017, 95, 060701.

6802 | Soft Matter, 2019, 15, 6791--6802 This journal is © The Royal Society of Chemistry 2019