Soft Matter

View Article Online

PAPER View Journal

A theory for the effect of patch/non-patch

Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

attractions on the self-assembly of patchy colloids

Cite this: DOI: 10.1039/c7sm01263b
Bennett D. Marshall

In this paper, we develop a thermodynamic perturbation theory to describe the self-assembly of patchy
colloids which exhibit both patch–patch attractions as well as patch/non-patch attractions. That is,
patches attract other patches as well as the no patch region (we call this region C). In general, the
patch–patch and patch–C attractions operate on different energy scales allowing for a competition between
Received 27th June 2017, different modes of attraction. This competition may result in anomalous thermodynamic properties. As an
Accepted 4th September 2017 application, we tune the patch parameters to reproduce the liquid density (suitably scaled) maximum of water.
DOI: 10.1039/c7sm01263b It is then shown that the liquid branch of the colloids phase diagram has liquid densities consistent with both
saturated and super-cooled liquid water. Finally, it is shown that the colloids reproduce water’s anomalous
rsc.li/soft-matter-journal minimum in isothermal compressibility and negative volume expansivity.

I Introduction
Self-assembly of patchy colloids has been the subject of extensive
experimental1–4 and theoretical5–8 investigation. Patchy colloids
(PC’s) are spherical colloids with discrete surface patches which
attract other patches, but have only repulsion between the Fig. 1 Diagram of interacting competitive colloids. The grey region is
patches and the non-patch region (the surface area not covered called C. Both patch–patch and patch–C attractions are allowed.
by patches). These patch–patch attractions are highly directional.
This anisotropy results in controlled self-assembly into prede-
signed structures as well as rich phase behavior. Another class of the 2A–9B colloids14 which exhibit a reentrant phase diagram
anisotropic colloids are inverse patchy colloids (IPC’s).9–13 As the are composed of two large A type patches in an axisymmetric
name suggest, IPC’s are the inverse of PC’s. There are a number location, and 9 smaller B type patches along the equator. While
of surface patches on the colloid, but patch–patch interactions well-defined from a theoretical stand point, synthesis of these
are repulsive instead of attractive, and patch/non-patch attractions would be quite challenging. Competitive colloids on the other
are attractive instead of repulsive. It has been demonstrated that hand may be comparatively easy to synthesize. It has been
IPC’s exhibit a rich variety of phase behavior. experimentally demonstrated that patchy colloids with several
What has not been studied (to the authors knowledge) is the patches can be synthesized.2 The patch–patch attractions are
class of colloids which exhibit both patch–patch and patch– mediated by DNA sticky ends. For the case of competitive colloids,
non-patch attractions. We will call the surface area of the colloid a judicious choice of patch DNA sticky ends and the DNA sequence
which is not a patch C, and will call this class of colloids, of C sticky ends would allow researchers to control the patch–patch
competitive colloids. See Fig. 1 for an illustration. An interesting binding energies as well as the patch–C binding energies.
feature of competitive colloids is that certain choices of patch Wertheim’s16,17 thermodynamic perturbation theory (TPT)
size and attractive energies could set up a competition between for associating fluids has proven to be a valuable tool in the
patch–patch and patch–C attractions resulting in exotic phase theoretical study of the self-assembly and phase behavior of
behavior such as a reentrant phase diagram. PC’s.6,14,18,19 Wertheim’s reformulation of statistical mechanics
Theoretical and Monte Carlo simulation calculations have into a multi-density formalism is designed to accommodate the
shown that certain classes of PC’s exhibit this type of phase strength and limited valence of association interactions. While
behavior.14,15 One drawback of these PC models is that the formally exact in its most general form, when applied as a
patchy colloids often require exotic patch designs. For instance, perturbation theory several assumptions are made. Perturbation
theory is defined by the neglect of all contributions to the change
ExxonMobil Research and Engineering, 22777 Springwoods Village Parkway, Spring, in free energy due to association which represent interactions
TX 77389, USA. E-mail: bennettd1980@gmail.com between distinct associated clusters. This allows the free energy

This journal is © The Royal Society of Chemistry 2017 Soft Matter

 View Article Online

Paper Soft Matter

to be described by reference system correlations. When applied at anomalous properties of water using TPT1 for colloidal models
first order (TPT1) it is assumed that all patches are independent based solely on the transition to tetrahedral symmetry. That is,
of one another (no patch–patch correlations) and all patches are in TPT1 the theory knows that a tetrahedral patchy colloid can
singly bondable. Going to higher order in perturbation allows the only be bonded to 4 other colloids; however, the theory has no
inclusion of steric effects between patches,20 rings20,21 and multiple knowledge that the patches are arranged in a tetrahedral
bonding of patches.22,23 In addition to perturbation theory, the arrangement. Smallenburg et al.35 showed that if one considers
more complex formalism of integral equation theory (IET) can non-additive hard sphere diameters that TPT1 could be used to
be employed.24,25 predict anomalous features of liquid water. However, as discussed
Given the remarkable success in applying TPT1 to describe above, it is the tetrahedral symmetry of fully bonded liquid water
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

the self-assembly and phase behavior of patchy colloid fluids. which results in the unique properties of water, not non-additive
It would be equally valuable to have such a tool to describe hard sphere diameters.
the self-assembly and phase behavior of competitive colloids. While TPT cannot strictly be used to predict the structural
Currently, a TPT has not been developed which could be anomalies of water due to tetrahedral symmetry, it is possible
applied to competitive colloid interaction. to use TPT as a tool to design patchy colloid models which
The first challenge in the development of such a theory is mimic the anomalous behavior of water by setting up a competition
the accounting for steric effects between colloids which are of energy scales of attraction.
simultaneously bonded to C on the same colloid. Previous Recently, Rovigatti et al.36 showed that patchy colloid models
perturbation theories19,22,26–28 have been developed to account which exhibit re-entrant phase behavior (2A–9B colloids14) could
for multiple bonding of patches. However, applicable forms of be used to explain the possible divergence of the compressibility
these theories have only been developed for patches which and specific heat in supercooled water.
bond up to twice. The C region on competitive colloids can As a first application of the competitive colloid model, we
receive as many association bonds as sterically allowed. This map the thermodynamic behavior of water onto a competitive
will, in general, allow for C to receive more than two association colloid potential of interaction. Going beyond qualitative comparison
bonds. A similar problem was addressed in the development of to provide an experimentally testable prediction, we fit competitive
a TPT for binary mixture of patchy and spherically symmetric colloid parameters (patch size and attractive energies) which
colloids where the patches on the patchy colloid can bond to reproduce the density of water (suitably scaled) in the tempera-
other patches or with the spherically symmetric colloid.18,29,30 ture range T = 0–100 1C. With the parameters fixed, the full
In the development of the TPT for competitive colloids, we treat phase diagram is predicted. We demonstrate how the density
steric hindrance for C in a similar fashion to the spherically maximum of water can be reproduced using a competitive
symmetric colloid in these binary mixtures. colloid model. At the point of the density maximum, patch–C
With the multiple bonding of C addressed, the challenge is association bonds are being traded for patch–patch association
then to develop a general thermodynamic equation of state bonds. It is also shown that, like liquid water, this results in a
which can be used for competitive colloids with any number minimum in the isothermal compressibility as a function of
and functionality of singly bonded patches. It is the generality temperature and negative volume expansivity.
of the TPT1 solution which has led to it’s widespread application.
Here, we seek a similar universal solution for competitive colloids.
The resulting free energy presented in section II is a very
general solution to TPT for the case that there are an arbitrary II Theory
number and functionality of singly bondable patches which can
associate with a large irregular C of arbitrary shape. The (A) Potential of interaction
number of association bonds which C can receive is dictated In this section, we describe a simple geometric form for the
by the geometry of C and the resulting steric effects between potential of interaction of spherical colloids of diameter d with
colloids bonded to C. Competitive colloids are one application anisotropic attractions. We allow for the colloids to have a set of
of this general solution. singly bondable attractive patches Gp = {A, B, C,. . ., Z}. These
The thermodynamics of liquid and super-cooled water has patches can attract each other (or not) as well as have attractions
been the subject of intense research for decades.31 The fascination to the surface of the colloid which is not covered by patches. We
with water stems from its anomalous thermodynamic properties call this no patch region C. The total set of association sites is
which result from the structural transition to tetrahedral then G = {C, Gp}.
coordination which accompanies water becoming fully hydrogen In TPT1 the size of all patches is restricted such that steric
bonded.32 There has also been significant interest in developing effects prevent a patch bonding more than once. For the
patchy colloid models which reproduce these anomalous features patches in Gp we enforce this single bonding condition. However,
of water. Sciortino and coworkers33,34 have extensively studied 4 in general, the C region may receive several association bonds
patch colloidal models with tetrahedral symmetry using molecular due to its larger size. For this reason, we do not restrict the
simulation. interaction between the patches and C. The maximum number
Unfortunately, since perturbation theory assumes an unchanging of association bonds of C is nmax which is set by steric repulsions.
hard sphere reference fluid, it is not possible to reproduce these This is illustrated in Fig. 2.

Soft Matter This journal is © The Royal Society of Chemistry 2017

 View Article Online

Soft Matter Paper

Fig. 2 Diagram depicting steric restrictions in patch–patch and patch–C Fig. 3 Diagram of conical square well patch–patch attraction. The patch
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

interactions. size is controlled the critical angle yc.

The potential of interaction is then given as the sum of hard correctly oriented for association yP1 o yc,P, but all patches
sphere + association contributions D of colloid 2 are not correctly oriented for association
X X X yD2 4 yc,D 8 D A Gp.
jð12Þ ¼ jHS ðr12 Þ þ jAB ð12Þ þ ðjCC ð12Þ þ jCC ð12ÞÞ Fig. 3 illustrates the interaction of two colloids. If r o rc with
A2Gp B2Gp C2Gp
yA o yc,A and yB o yc,B a patch–patch association bond is
(1) formed. However, if r o rc with yA o yc,A and yB 4 yc,B a patch–
C bond is formed. The potential eqn (1)–(4) is general enough
The term jHS(r) is the potential of a hard sphere fluid
( to allow for the inverse patchy colloid type of attraction as well
1 rod as allow one to probe the effects of patch–patch attractions and
jHS ðrÞ ¼ (2)
patch size on the phase behavior of this class of colloids. We will
0 Otherwise
call this class of colloids competitive colloids. In what follows
In eqn (1), jAB(12) represents the association potential between we apply thermodynamic perturbation theory to develop an
patch A on colloid 1 and patch B on colloid 2, and jCC (12) analytical model to describe competitive colloids.
represents the association potential between patch C on colloid
1 and non-patch surface C on colloid 2. The notation (1)  (B) Thermodynamic perturbation theory
- -
(r1,O1) represents the position r1 and orientation O1 of colloid 1 We develop the theory in the multi-density formalism of
and r12 is the distance between centers of the pair. Wertheim16,17,39 where each bonding state of a colloid is
As a primitive model for the patch–patch potential we assigned a number density. The density of colloids bonded at the
employ conical square well association sites (also known as set of sites a is given by ra. To aid in the reduction to irreducible
the Kern–Frenkel model of PC’s)37,38 graphs, Wertheim introduced the density parameters sg
( X
eAB r12  rc and yA1  yc;A and yB2  yc;B sg ¼ ra (5)
jAB ð12Þ ¼ ag
0 otherwise
(3) where the empty set a = + is included in the sum. Two notable
cases of eqn (5) are sG = r and so = ro; where r is the total number
where rc is the maximum distance between colloids for which density of colloids and ro is the density of colloids not bonded at
association can occur, yA1 is the angle between the center of any patch including C (monomer density).
patch A on colloid 1 and the vector connecting the two centers, In Wertheim’s multi-density formalism, the exact change in
and yc is the maximum angle for which association can occur. free energy due to association is given by
With this, if two colloids are both positioned and oriented  
A  AHS AAS r .
correctly, a bond is formed and the energy of the system is ¼ ¼ r ln o þ Q þ r  DcðoÞ V (6)
decreased by a factor eAB. Here we assume a constant value of rc, VkB T VkB T r
but allow yc to vary as long as it is small enough to ensure the
where V is the system volume, kB is Boltzmann’s constant, T is
single bond per patch condition.
temperature, AHS is the hard sphere reference free energy and Q
Similarly, the patch–C potential is given as the orientationally
is given by
dependent square well contribution X
( Q ¼ r þ cg sGg (7)
ePC r12  rc and yP1 o yc;P and yD2 4 yc;D 8 D 2 Gp
gG
jPC ð12Þ ¼ ga+
0 otherwise
(4) The term Dc(o) is the associative contribution to the funda-
mental graph sum which encodes all association attractions
Eqn (4) states that there is a square well attraction of depth between the colloids, and cg is obtained from the relation
ePC between the patch P on colloid 1 and the non-patch surface 
C on colloid 2 if the colloids centers are within a distance rc, @DcðoÞ V
cg ¼ (8)
the orientations are such that the patch P of colloid 1 is @sGg

This journal is © The Royal Society of Chemistry 2017 Soft Matter

 View Article Online

Paper Soft Matter

where in eqn (8) g a +. The graph sum Dc(o) is decomposed as Now that Dc(o) has been fully specified, the densities of the
various bonding states can be calculated through the relation27
Dc(o) = Dc(o) (o)
pp + DcpC (9)
rg X Y
where Dc(o) ¼ ct (15)
pp accounts for the attractions between patches in Gp ro PðgÞ¼ftg t
(o)
and DcpC accounts for attractions between patches in Gp and C.
We treat the interaction between Gp patches in first order where P(g) is the partition of the set g into non-empty subsets.
perturbation theory (TPT1) giving Dc(o)pp as
40
For example, the density rABC is given by rABC = ro(cABC + cABcC +
. xX X cBCcA + cCAcB + cAcBcC). We note the following relation holds
DcðoÞ
pp V ¼ sGA kAB fAB sGB (10)
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

2 A2G B2Gp cg = 0 for n(g) 4 1 (16)

p

This allows eqn (15) to be simplified to

In eqn (10) kAB = (1  cos yc,A)(1  cos yc,B)/4 is the probability
that two colloids are oriented such that patch A on colloid 1 can rg Y rA
¼ (17)
bond to patch B on colloid 2. Also, ro A2g ro
ð rc
x ¼ 4p r2 yHS ðrÞdr (11) The relation in (17) implies40 the following relations for the
d fraction of colloids not bonded at patch D
is the integral of the hard sphere reference cavity correlation sGD 1
XD ¼ ¼ (18)
function over the bond volume. For this quantity we employ r 1 þ cD
the method of Marshall and Chapman.20 Lastly, the term
fAB = exp(eAB/kBT)  1 is the magnitude of the association Mayer as well as the relation for the monomer fraction
function. Y
Xo ¼ XD (19)
For the attractions between Gp patches and C we must A2G
explicitly account for the fact that the C (non-patch) region of
the colloid can bond up to a maximum of nmax times. To The term cD is obtained through (8). For the case D A Gp
accomplish this, we take a similar approach to that taken for X nmax
X n pffiffiffiffiffiffiffiffiffi
mixtures of patchy and spherically symmetric colloids.18,29 The cD ¼ xr kAD fAD XA þ rXC yHS ðdÞ kDD fDC Dn1 XðnÞ dðnÞ
n!
general form of the graph sum for this interaction remains A2Gp n¼1

unchanged (eqn (13) and (14) of ref. 18), except for the change (20)
that the current approach is for a pure component fluid which
and for D = C
o - sG–C = sGp to obtain
results in the change r(s)
.
ðoÞ
nmax Dc
X pC V
ðoÞ nmax
DcpC X 1 n ðnÞ ðnÞ 1 n ðnÞ ðnÞ
¼ sGC D d X (12) cC ¼ D d X ¼ (21)
V n! n¼1
n! sGp
n¼1

In eqn (12) n is the number of incident association bonds on C We now define the fraction of colloids with C bonded n times
and D is (irrespective of bonding of the patches in Gp) as wC,n. These
X fractions must sum to unity
pffiffiffiffiffiffiffiffi
D ¼ yHS ðdÞ sGL fLC kLL (13)
nmax
X
L2Gp
wC;n ¼ 1 (22)
n¼0
d(n) is the second order correction to the first order super-
position of the many body correlation function for the associated Note that wC,o = XC is simply the fraction of colloids not bonded
cluster. This term is evaluated using the branched TPT2 solution at C. Comparing eqn (18), (21) and (22) we deduce the fractions
of Marshall and Chapman41 as for n 4 0
8  pffiffiffiffiffiffiffiffiffiffiffiffiffiffi3
> 1
>
< ð1 þ 4lÞn3 1 þ 1 þ 4l wC;n ¼ w Dn dðnÞ XðnÞ for n 4 0 (23)
2 for n 4 1 n! C;o
dðnÞ ¼ 2 (14)
>
>
: From eqn (7)
1 for n ¼ 1
Q X
The term l = 0.2336Z + 0.1067Z2 where Z is the packing fraction.42 ¼ 1 þ c A XA (24)
r A2G
Finally, the terms X(n) are the cluster partition functions which
are independent of temperature and density. These terms encode which when combined with eqn (18) gives
the steric effects rich arise when multiple colloids attempt to
Q X
donate an association bond to C. We will discuss the cluster ¼ 1 þ ð1  XA Þ (25)
r
partition functions in Section C. A2G

Soft Matter This journal is © The Royal Society of Chemistry 2017

 View Article Online

Soft Matter Paper

The patch–patch contribution to the graph sum is found to be Combining (30)–(32)

ðoÞ nmax
0 1
Dcpp rX rX rX r mAS X r@ X @ ln x
¼ XA cA  nwC;n ¼ ð1  XA Þ  nC ¼ ln XA  ð1  XA Þ  nC A
V 2 A2G 2 n¼1 2 A2G 2 kB T 2 @r
p p A2G A2G p

(26) (33)
max
!
@ ln yHS ðdÞ X
n
@ ln dðnÞ
where n% C is the average number of incident association bonds on C r nC þ wC;n
@r n¼1
@r
max
n
X
nC ¼ nwC;n (27) The bonding state of the colloid is defined through the
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

n¼0
fractions XA from which the fractions wC,n can be evaluated. For
From eqn (12), (22) and (23) a colloid with np patches in Gp, this in general requires the
simultaneous solution of np + 1 equations defined by eqn (18)
ðoÞ nmax
DcpC X   with the quantities cD defined by eqn (20) and (21).
¼r wC;n ¼ r 1  wC;o ¼ rð1  XC Þ (28)
V n¼1
(C) Cluster partition functions
Combining these results, we obtain the association contribution The perturbation theory derived in Section B is a general result.
to the free energy (N – total number of colloids) Specification to the specific competitive colloid potential
AAS X XA 1

nC defined through eqn (1)–(4) is achieved through evaluation of
¼ ln XA  þ þ ln XC þ (29) the cluster partition functions X(n). Here n is the number of
NkB T A2G 2 2 2
p
colloids with patches bonded to C on a center colloid in the
The contribution within the summation of eqn (29) gives the cluster. The cluster partition functions are related to the
standard TPT1 free energy for the patches in Gp, while the number of states for which a given associated cluster can exist.
remaining terms represent corrections for the fact that C can To derive these functions, we modify the approach previously
receive multiple association bonds. developed to calculate X(n) for mixtures of patchy and spherically
Eqn (29) is a very general solution to TPT for the case that symmetric colloids.29
there are an arbitrary number and functionality of singly bondable To proceed we consider a colloid which we label 1. This
patches in Gp which can associate with a large irregular C of colloid is bonded at C to patches on n additional colloids.
arbitrary shape and functionality which can receive as many These n colloids are labelled from 2 to n + 1. Now we integrate
association bonds as are sterically allowed. This multi-body steric the position of the n colloids over all possible bonding states of
hindrance is encoded in the single cluster partition function X(n). this cluster. We evaluate this integral with the mean value
While these steric effects are included for multiple bonding of C, theorem as
correlations between separate patches are neglected and treated in X(n) = vnCP(n) (34)
first order perturbation theory. This approach allows for a relatively
(n)
simple and elegant solution, while including full steric effects on Where P is the probability that n colloids can be generated in
the multiple bonding of C. the bond volume of C with no hard sphere overlap. The bond
The chemical potential is obtained using the general volume vC is given as the surface area of the colloid which is
relation16 devoid of patches integrated from the hard sphere diameter d
to the critical radius rc
mAS m  mHS @ DcðoÞ
¼ ¼ ln Xo  (30) 0 1
kB T kB T @r V X 
rc3  d 3 @
vC ¼ 4p  2p 1  cos yc;A A (35)
Taking the derivative of the individual contributions 3 A2G p
0 1
@ Dcpp r@ X @ ln x For the probability P(n) we employ the single cluster probabil-
¼ ð1  XA Þ  nC A (31)
@r V 2 A2G
p
@r ities calculated in ref. 29. That is, we assume that P(n) can be
approximated as the value obtained for the case that the center
and colloid 1 has no patches. We have not neglected the fact that
max the patches decrease the available surface area for patch–C
@ DcpC @ ln yHS ðdÞ X
n
n
¼r wC;o Dn dðnÞ XðnÞ attractions, this effect is included in the bond volume eqn (35).
@r V @r n¼1
n! This should be a reasonable approximation for competitive
nmax
colloids since the maximum average number of bonds C could
X 1 @ ln dðnÞ
þr wC;o Dn dðnÞ XðnÞ (32) receive would be equal to the number of attractive patches. As
n¼1
n! @r patchy colloids typically have 1–5 patches, we should not expect
nmax strong competition for association at C as with mixtures of
@ ln yHS ðdÞ X @ ln dðnÞ
¼r nC þ r wC;n spherically symmetric and patchy colloids when the spherically
@r @r
n¼1 symmetric colloids are dilute.18,29,30

This journal is © The Royal Society of Chemistry 2017 Soft Matter

 View Article Online

Paper Soft Matter

(D) Competitive colloids with k equivalent patches patch size at yc o 301 such that we can guarantee that the patches
We now specialize the general approach developed in Section B are singly bondable, hence satisfying the constraints of the theory.
to the specific case of competitive colloids defined as having l We do not define the colloidal diameter d as it will simply apply a
equivalent monovalent patches A with patch–patch attractions scaling to our results. The parameters eAA, eAC and yc are then
given by the energy scale eAA. In addition to patch–patch adjusted to minimize the error between the scaled saturated liquid
attractions, the A patches are also attracted to the non-patch densities of the colloid and the corresponding densities for
C region of the colloid with an energy scale eAC. The ratio RAC water in the temperature range 0 1C o T o 100 1C. We make
defined in eqn (36) will then quantify the strength of AA comparisons in terms of scaled variables, due to the fact the
attractions in relation to AC attractions diameter of water is on the order of angstroms and colloids
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

would likely be on the order of nanometers to microns. We

eAA
RAC ¼ (36) consider the following reduced densities
eAC
r* = rd3 (41)
To solve for the bonding fractions XA we combine eqn (18),
(20), (22) and (23) to obtain the following closed equation for XA Neutron diffraction43 shows that the first maximum in oxygen–
nmax
P n pffiffiffiffiffiffiffiffi oxygen correlation function in liquid water is located at a
rXA yHS ðdÞ kAA fAC Dn1 XðnÞ dðnÞ distance of 2.75 Å. SPC water44 uses a diameter of 3 Å. In this
n¼1 n!
XA þ lxrkAA fAA XA2 þ max
nP 1
¼ 1 work we determined a diameter of 2.85 Å gave the best agree-
1þ Dn dðnÞ XðnÞ ment with experiment. Hence, when we compare to scaled
n¼1 n!
water properties, we have scaled using a water diameter of
(37)
dw = 2.85 Å.
with We attempted patch numbers 2 r l r 4, but the l = 4 case is
pffiffiffiffiffiffiffiffi the only one for which we were able to reproduce the scaled
D ¼ lyHS ðdÞrXA fAC kAA (38)
densities of liquid water 0 1C o T o 100 1C. Fig. 4 compares
Eqn (37) can be interpreted as the sum over the bonding states scaled saturated liquid densities for water and our 4-patch
of the A patches. The first term on the left-hand side is the competitive colloid. The patch parameters are given in Table 1.
fraction of patches which are not bonded, the center term is the Overall, the colloidal model is able to satisfactorily describe the
fraction of A patches which are bonded to other A patches wAA, saturated liquid densities of water over this temperature range.
and finally the last term is the fraction of A patches which are In Fig. 5 we calculate the temperature–density phase dia-
bonded to the no patch C region wAC. gram for the competitive colloid and compare to both saturated
wAA ¼ lxrkAA fAA XA2 ; liquid densities of water in the temperature range 273 K o T o
573 K and the liquid phase densities of super-cooled water in
max
nP n pffiffiffiffiffiffiffiffi the temperature range 130 K o T o 273 K. Both the competitive
rXA yHS ðdÞ kAA fAC Dn1 XðnÞ dðnÞ (39)
n¼1 n! colloid as well as liquid water show a decreasing density with
wAC ¼ max
nP 1 n ðnÞ ðnÞ decreasing T in the super-cooled region. However, the competitive
1þ D d X
n¼1 n! colloid has a monotonically decreasing density in this region. This
is not the case for water which exhibits a density minimum
A quantity which describes the competition between AA and
AC attractions is Y which is defined as the ratio
w
Y ¼ AA (40)
wAC

For Y 4 1, AA interactions dominate while for Y o 1, AC

interactions dominate. Unlike eqn (36), Y depends on density
and temperature. Upon solution of eqn (37) for XA the fractions
XC can be solved for using eqn (18) and (21).

III Application to liquid water

Fig. 4 Competitive colloid model predictions (curve) and experimental
In this section, we apply the theory developed in II to develop a data (symbols)45 for the scaled saturated densities of liquid water.
competitive colloid model which partially reproduces the
anomalous properties of liquid water. The model is described
by the number of patches l, the size of the patches yc, critical
Table 1 Competitive colloid model parameters which reproduce the
radius rc, the patch–patch well depth eAA and the patch–C well scaled liquid density of water at ambient conditions
depth eAC as well as the colloidal diameter d. For the critical radius
we assume a value rc = 1.1d which is consistent with previous Patches rc yc eAA/kB (K) RAC
experimental realizations1,2 of patchy colloids. We bound the 4 1.1d 22.241 3067.8 2.183

Soft Matter This journal is © The Royal Society of Chemistry 2017

 View Article Online

Soft Matter Paper

Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

Fig. 5 Model predictions (curve) of the phase diagram of competitive Fig. 7 Fraction of colloids with C bonded n times for a saturated liquid at
colloids compared to saturated liquid density data45 (solid circles) and T = 300 K.
super-cooled density data46 (open circles) of liquid water.

near T = 200 K. Finally, we note the critical temperature of the

competitive colloid Tc B 534 K is substantially lower than water
Tc = 647 K for water. This difference is a result of the fact that
unlike water, the competitive colloids only exhibit short ranged
attractions.
Fig. 6 shows the total average number of bonded patches per
colloid, average number of patches participating in patch–C
bonds per colloid, as well as the average number of patch–patch
bonds per colloid. Starting from a high T = 534 K near the
critical point, the number of patch–C bonds increases as
Fig. 8 Comparison of competitive colloid model predictions for the
temperature is decreased. Decreasing temperature results in scaled isothermal compressibility of a saturated liquid phase (curve) to
an increase in the number of patch–patch bonds, eventually experimental data49 for water (symbols).
forcing a maximum in the number of patch–C bonds near
T = 315 K. Fig. 7 shows the distribution of wC,n (eqn (23)) in the
saturated liquid at T = 300 K. As can be seen, there is a non- model predictions for a saturated liquid. As can be seen, both the
negligible contribution from colloids which have C bonded up competitive colloid and water have their respective minima at
to 5 times. Further decrease in temperature results in an T = 319 K. The two scaled data sets are in very good agreement
increase in the number of patch–patch bonds, and a decrease for T 4 273 K. In the super-cooled water regime (T o 273 K) the
in patch–C bonds. It is this switching from patch–C bonds to k of water increases more rapidly upon cooling than the
patch–patch bonds which results in the density maximum. competitive colloid. Like the density maximum, the minimum
The transition of water towards a tetrahedral hydrogen bonding in k for the competitive colloid is a result of switching patch–C
network results in a minimum in the isothermal compressibility k bonds to patch–patch bonds as temperature is lowered.
of liquid water.47,48 In Fig. 8 we compare k* = k/k(303 K) versus T Finally, Fig. 9 compares the volume expansivity a = q ln r/qT
for liquid phase and supercooled water to the competitive colloid of liquid water at ambient conditions to the expansivity predicted
for a saturated liquid by the competitive colloid model. Model
and experiment are in good agreement for temperatures T 4
270 K. The competitive colloid model successfully reproduces
the anomalous feature of a o 0.

Fig. 6 Competitive colloid model predictions for the average number of

bonded patches per colloid (long dashed curve), number of patch–patch
bonds per colloid (short dashed curve) and average number of patches
participating in patch–C bonds per colloid (solid curve) for a saturated Fig. 9 Comparison of experiment (symbols)50 and competitive colloid
liquid. predictions (curve) for the volume expansivity of liquid water.

This journal is © The Royal Society of Chemistry 2017 Soft Matter

 View Article Online

Paper Soft Matter

IV Discussion and summary 12 P. D. J. van Oostrum, M. Hejazifar, C. Niedermayer and

E. Reimhult, J. Phys.: Condens. Matter, 2015, 27, 234105.
We have developed a thermodynamic perturbation theory to 13 E. Bianchi, G. Kahl, C. N. Likos, P. Pincus, M. Sztucki,
describe the self-assembly and phase behavior of competitive A. Moussaid, T. Narayanan and F. Sciortino, Soft Matter,
colloids. As with previous models of patchy colloids,14,15,36 2011, 7, 8313.
competition between different modes of attraction results in 14 J. Russo, J. M. Tavares, P. I. C. Teixeira, M. M. Telo da Gama
anomalous phase behavior. These previous patchy colloid and F. Sciortino, J. Chem. Phys., 2011, 135, 34501.
models required two types of patches (bi-functional) to obtain 15 Y. V. Kalyuzhnyi and P. T. Cummings, J. Chem. Phys., 2013,
competing energy scales. An advantage of competitive colloids 139, 104905.
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

is the plausibility of laboratory synthesis. It has been previously 16 M. S. Wertheim, J. Stat. Phys., 1986, 42, 459.
demonstrated that 4 patch colloids, with a single type of patch, 17 M. S. Wertheim, J. Stat. Phys., 1986, 42, 477.
can indeed be synthesized.2 18 B. D. Marshall and W. G. Chapman, Soft Matter, 2013, 9, 11346.
As an application of the new theory, we adjusted the patch 19 A. Haghmoradi, L. Wang and W. G. Chapman, J. Phys.:
parameters to reproduce the density maximum of liquid water Condens. Matter, 2016, 28, 244009.
(appropriately scaled). Unlike water which has 2 association 20 B. D. Marshall and W. G. Chapman, Phys. Rev. E, 2013,
donor and 2 acceptor sites, we chose 4 equivalent patches for 87, 52307.
ease of laboratory synthesis. It was shown that the theory also 21 J. M. Tavares, N. G. Almarza, M. M. Telo da Gama, P. Tartaglia,
predicted anomalous features of thermodynamic functions F. Sciortino, F. Sciortino, T. Bellini, E. Bourgeat-Lami, E. Duguet
such as the minimum in the compressibility, consistent with and S. Ravaine, Soft Matter, 2015, 11, 5828.
liquid water. The anomalous properties of water are the result 22 Y. V. Kalyuzhnyi, B. D. Marshall, W. G. Chapman and
of the transition from normal fluid behavior at high temperatures P. T. Cummings, J. Chem. Phys., 2013, 139, 044909.
to tetrahedral structure at ambient and super-cooled temperatures. 23 B. D. Marshall, D. Ballal and W. G. Chapman, J. Chem. Phys.,
The anomalous behavior of the competitive colloids is the 2012, 137, 104909.
result of the trading of patch–C bonds to patch–patch bonds 24 Y. V. Kalyuzhnyi, V. Vlachy, M. F. Holovko and G. Stell,
as temperature is lowered. J. Chem. Phys., 1995, 102, 5770.
25 Y. V. Kalyuzhnyi and G. Stell, Mol. Phys., 1993, 78, 1247.
26 Y. V. Kalyuzhnyi, H. Docherty and P. T. Cummings, J. Chem.
Conflicts of interest Phys., 2011, 135, 14501.
There are no conflicts to declare. 27 B. D. Marshall, D. Ballal and W. G. Chapman, J. Chem. Phys.,
2012, 137, 104909.
28 B. D. Marshall and W. G. Chapman, Soft Matter, 2014,
10, 5168.
References
29 B. D. Marshall and W. G. Chapman, J. Chem. Phys., 2013,
1 L. Feng, R. Dreyfus, R. Sha, N. C. Seeman and P. M. Chaikin, 139, 104904.
Adv. Mater., 2013, 25, 2779. 30 A. Bansal, D. Asthagiri, K. R. Cox and W. G. Chapman,
2 Y. Wang, Y. Wang, D. R. Breed, V. N. Manoharan, L. Feng, J. Chem. Phys., 2016, 145, 74904.
A. D. Hollingsworth, M. Weck and D. J. Pine, Nature, 2012, 31 P. Gallo, K. Amann-Winkel, C. A. Angell, M. A. Anisimov,
491, 51. F. Caupin, C. Chakravarty, E. Lascaris, T. Loerting,
3 A. B. Pawar and I. Kretzschmar, Langmuir, 2008, 24, 355. A. Z. Panagiotopoulos, J. Russo, J. A. Sellberg, H. E. Stanley,
4 A. B. Pawar and I. Kretzschmar, Macromol. Rapid Commun., H. Tanaka, C. Vega, L. Xu and L. G. M. Pettersson, Chem. Rev.,
2010, 31, 150. 2016, 116, 7463.
5 F. Sciortino, E. Bianchi, J. F. Douglas and P. Tartaglia, 32 J. R. Errington and P. G. Debenedetti, Nature, 2001, 409, 318.
J. Chem. Phys., 2007, 126, 194903. 33 I. Saika-Voivod, F. Smallenburg and F. Sciortino, J. Chem.
6 E. Bianchi, R. Blaak and C. N. Likos, Phys. Chem. Chem. Phys., 2013, 139, 234901.
Phys., 2011, 13, 6397. 34 F. Romano, P. Tartaglia and F. Sciortino, J. Phys.: Condens.
7 Z. Preisler, T. Vissers, F. Smallenburg, G. Munaò and Matter, 2007, 19, 322101.
F. Sciortino, J. Phys. Chem. B, 2013, 117, 9540. 35 F. Smallenburg, L. Filion and F. Sciortino, J. Phys. Chem. B,
8 Z. Zhang, A. S. Keys, T. Chen and S. C. Glotzer, Langmuir, 2015, 119, 9076.
2005, 21, 11547. 36 L. Rovigatti, V. Bianco, J. M. Tavares and F. Sciortino,
9 E. G. Noya, I. Kolovos, G. Doppelbauer, G. Kahl, E. Bianchi, J. Chem. Phys., 2017, 146, 41103.
A. W. Wilber, H. C. Kok, R. Lyus, D. C. Lee, C. Klinke and 37 N. Kern and D. Frenkel, J. Chem. Phys., 2003, 118, 9882.
H. Weller, Soft Matter, 2014, 10, 8464. 38 G. Jackson, W. G. Chapman and K. E. Gubbins, Mol. Phys.,
10 Y. V. Kalyuzhnyi, E. Bianchi, S. Ferrari and G. Kahl, J. Chem. 1988, 65, 1.
Phys., 2015, 142, 114108. 39 M. S. Wertheim, J. Chem. Phys., 1987, 87, 7323.
11 E. G. Noya and E. Bianchi, J. Phys.: Condens. Matter, 2015, 40 W. G. Chapman, PhD dissertation, Cornell University,
27, 234103. Ithaca, NY, 1988.

Soft Matter This journal is © The Royal Society of Chemistry 2017

 View Article Online

Soft Matter Paper

41 B. D. Marshall and W. G. Chapman, J. Chem. Phys., 2013, 45 G. Haar and G. Kell, NBS/NRC STEAM TABLES, Hemisphere
138, 174109. Publishing, Washington DC, 1984.
42 S. Phan, E. Kierlik, M. L. Rosinberg, H. Yu and G. Stell, 46 F. Mallamace, C. Branca, M. Broccio, C. Corsaro, C.-Y. Mou and
J. Chem. Phys., 1993, 99, 5326. S.-H. Chen, Proc. Natl. Acad. Sci. U. S. A., 2007, 104, 18387.
43 A. K. Soper, F. Bruni and M. A. Ricci, J. Chem. Phys., 1998, 47 P. G. Debenedetti, J. Phys.: Condens. Matter, 2003, 15, R1669.
106, 247. 48 R. J. Speedy and C. A. Angell, J. Chem. Phys., 1976, 65, 851.
44 H. J. Berdensen, J. R. Grigera and T. P. Straatsma, J. Chem. 49 S. Kell, J. Chem. Eng. Data, 1975, 20, 97.
Phys., 1987, 91, 6269. 50 G. S. Kell, J. Chem. Eng. Data, 1967, 12, 66.
Published on 11 September 2017. Downloaded by University of Windsor on 11/09/2017 09:32:55.

This journal is © The Royal Society of Chemistry 2017 Soft Matter