16. L. R. Moore, S. W. Chisholm, Limnol. Oceanogr. 44, 628 25. K. K. Cavender-Bares et al., Deep-Sea Res. Part I (to Z.I.J.

(to Z.I.J.), and NSF and the University of Tennessee (to

(1999). Oceanogr. Res. Pap. 48, 2373 (2001). E.R.Z.). It was also supported by the UK Natural
17. L. R. Moore et al., Limnol. Oceanogr. 47, 989 (2002). 26. L. R. Moore et al., Aquat. Microb. Ecol. 39, 257 (2005). Environment Research Council through the AMT consor-
18. E. Zinser et al., Appl. Environ. Microbiol. 72, 723 27. M. B. Sullivan et al., Nature 424, 1047 (2003). tium (NER/O/S/2001/00680). This is School of Ocean and
(2006). 28. B. C. Monger et al., Limnol. Oceanogr. 44, 1917 (1999). Earth Science and Technology contribution no. 6686 and
19. N. Ahlgren et al., Environ. Microbiol. 8, 441 (2006). 29. J. C. Venter et al., Science 304, 66 (2004). AMT contribution no. 107.
20. N. J. West, D. J. Scanlan, Appl. Environ. Microbiol. 65, 30. The authors acknowledge C. Robinson and the captain
2585 (1999). and crew of Royal Research Ship James Clark Ross for our Supporting Online Material
21. J. Aiken et al., Prog. Oceanogr. 45, 257 (2000). participation in the AMT cruise and D. Veneziano for www.sciencemag.org/cgi/content/full/311/5768/1737/DC1
22. S. B. Hooker et al., Prog. Oceanogr. 45, 313 (2000). guidance with the statistical analysis. This work was Materials and Methods
23. M. S. Rappe, S. J. Giovannoni, Annu. Rev. Microbiol. 57, funded in part by the Gordon and Betty Moore Figs. S1 to S6
369 (2003). Foundation, the Seaver Foundation, NSF Biological
24. L. Campbell et al., Deep-Sea Res. Part II Top. Stud. Oceanography Division (to S.W.C.), and U.S. Department 27 July 2005; accepted 11 January 2006
Oceanogr. 45, 2301 (1998). of Energy (to S.W.C.), NOAA and the University of Hawaii 10.1126/science.1118052

REPORTS
General Strategies for caused by nanoparticle dispersion and the stan-
dard mixing entropy, is used to reconcile the
experimental observations and emphasize the
Nanoparticle Dispersion importance of the nanoparticle-to-polymer size
ratio in controlling nanoparticle dispersion.
Michael E. Mackay,1,2* Anish Tuteja,1 Phillip M. Duxbury,2 Craig J. Hawker,3,4 First we discuss the dispersion of fullerenes
Brooke Van Horn,4 Zhibin Guan,5 Guanghui Chen,5 R. S. Krishnan1 into polystyrene, motivated by earlier work that
suggested that fullerene dispersion in polymers
Traditionally the dispersion of particles in polymeric materials has proven difficult and (11) is poor, limiting their utility, for example,
frequently results in phase separation and agglomeration. We show that thermodynamically stable in solar cells (8, 9). For a polymer blend, the
dispersion of nanoparticles into a polymeric liquid is enhanced for systems where the radius of insertion energy of a linear polymer chain with
gyration of the linear polymer is greater than the radius of the nanoparticle. Dispersed another controls dispersion and grows with the
nanoparticles swell the linear polymer chains, resulting in a polymer radius of gyration that grows number of monomers in the chain. So, the in-
with the nanoparticle volume fraction. It is proposed that this entropically unfavorable process is sertion enthalpy of a chain of N monomers is
offset by an enthalpy gain due to an increase in molecular contacts at dispersed nanoparticle proportional to Nc, where c is the Flory mixing
surfaces as compared with the surfaces of phase-separated nanoparticles. Even when the dispersed parameter and is the primary cause of phase
state is thermodynamically stable, it may be inaccessible unless the correct processing strategy is separation in incompatible blends. Nanoparti-
adopted, which is particularly important for the case of fullerene dispersion into linear polymers. cles have an insertion enthalpy that grows in
proportion to the surface area of the nano-
olymer phase stability in solution (1) or quire nanoparticle dispersion, such as in the particle, yielding an insertion enthalpy of s È

P with another polymer (2) has been studied

for over 50 years and found to be a del-
icate balance of entropic and enthalpic contribu-
use of fullerenes to enhance the efficiency of
polymer-based photovoltaic devices (8, 9) and in
the control of polymer viscosity using nano-
Ac, where A 0 4pa2 for a nanoparticle of radius
a. Although this enhancement is not as strong
as for polymer blends, dispersion of nanopar-
tions to the total free energy. For example, it is particles (10). ticles still depends critically on c. We observed
possible to fractionate a polymer by size with a We demonstrate strategies for control of experimentally that it is possible to disperse
small change in solvent quality (3) and to con- nanoparticle dispersion in linear polymer melts. up to a concentration of 2 volume % of C60
trol miscibility of chemically identical polymers We start with discussion of processing proce- in linear, monodisperse polystyrene. At small
whose only difference is architecture (branch- dures that enable stable dispersion of fullerenes nanoparticle concentration, Flory theory (1)
ing) (4). More recently, the phase stability of and then present an experimental character- gives a binodal or phase stability volume frac-
nanoparticle-polymer blends has attracted in- ization of the parameters that control the
tense scrutiny (5) and is challenging to predict phase boundary between the dispersed and
because of computational difficulty in accessing phase-segregated states of carefully considered
the relevant length and time scales. Flory the- nanoparticle-polymer mixtures. Moreover, it has
ories, density functional theories, and molecular been proven possible to disperse polyethylene
dynamics methods provide essential guidance, nanoparticles in polystyrene despite the fact that
although accurate calculations are restricted to linear polyethylene–linear polystyrene is a clas-
two or at most a few nanoparticles in the relevant sic phase-separating blend, which implies that
size regime (6, 7). Despite these difficulties, a nanoparticle morphology may actually enhance
vast array of applications are emerging that re- dispersion. This hypothesis is tested by using a
system composed of polystyrene nanoparticles Fig. 1. (A) Rapid precipitation of fullerene-
1
Department of Chemical Engineering and Materials dispersed in linear polystyrene, because the polystyrene blends, followed by drying and melt
Science, 2Department of Physics and Astronomy, Michigan monomer-monomer contacts in this system are processing, allows manufacture of fibers. The
State University, East Lansing, MI 48824, USA. 3Materials the same for all of its constituents. An enthalpic
Research Laboratory, University of California, Santa
fibers contains 1 wt % C60 fullerenes that were
mechanism that arises from nanoparticle packing melt spun into long fibers with a diameter of
Barbara, CA 93106, USA. 4IBM Almaden Research Center,
650 Harry Road, San Jose, CA 95120, USA. 5Department of effects operates at the nanoscale and is necessary circa 1 mm. (B) Fullerene (1 wt %)-polystyrene
Chemistry, University of California, Irvine, CA 92697, USA. in order to understand dispersion in this size re- blends developed through regular solvent evap-
*To whom correspondence should be addressed. E-mail: gime. A Flory theory, which includes this en- oration produce large, phase-separated domains,
mackay@msu.edu thalpic contribution as well as chain stretching which are not apparent in the fiber.

1740 24 MARCH 2006 VOL 311 SCIENCE www.sciencemag.org

 REPORTS
tion (fB ) of fB , expE–(1 þ s)^, assuming the have an unfavorable mixing enthalpy and are A particularly clear illustration of the impor-
phase-separated fullerenes form a pure nano- a classic phase-separating system, with com- tance of the ratio a/Rg on nanoparticle dispersion
particle phase. The use of the experimental plete phase separation occurring at molecular is provided by a mixture consisting of cross-linked
value of fB 0 0.02 yields an insertion enthalpy weights typical of those used here. We have polystyrene nanoparticles (22) blended with linear
per fullerene on the order of s , 3. The molecular taken transmission electron microscope (TEM) polystyrene. We observed phase separation when
insertion energy per monomer (e) is given by s 0 images and collected small angle neutron the nanoparticle size was greater than the poly-
ze/kBT, where z is the coordination number; kB, scattering (SANS) data for a wide variety of mer radius of gyration in a manner similar to that
the Boltzmann constant; and T, the temperature; mixtures. The TEM image (Fig. 2B) illustrates observed in the case of polyethylene nanoparticle
yielding an insertion energy of e , 0.02 eV for the dispersion of dendritic polyethylene nano- mixtures discussed above.
fullerenes in polystyrene. This relatively small particles in 393-kD linear polystyrene, from Tightly cross-linked polystyrene nanoparti-
energy may be rationalized by the fact that which a nanoparticle radius of 10 to 15 nm can cles, where every fifth monomer unit is poten-
favorable molecular contacts between the aro- be extracted. Moreover, a Guinier analysis of tially cross-linked (22), were blended with linear
matic rings on polystyrene and the hexagons on SANS data (16, 17) from polyethylene nano- polystyrene (10). This system produces a stable
the surface of C60 may occur. particles in dilute solution yields a polyethylene- blend even when the interparticle gap reaches
Fullerene dispersion is enabled by using our nanoparticle radius of 12.8 T 0.1 nm, which is surprisingly small distances, suggesting the lin-
technique of rapidly precipitating the compo- consistent with the TEM measurement. Neutron- ear polystyrene molecule is highly distorted (10).
nents in a mutual nonsolvent (10, 12) to arrive scattering data for the same nanoparticles The distortion was directly measured through a
at a dried powder that is then thermally aged, blended with linear polystyrene melts with SANS Guinier analysis from a sample in which
allowing melt processing and fiber spinning different molecular weights are also presented 2 weight (wt) % deuterated linear polystyrene
(Fig. 1A). It is known that fullerenes have lim- (Fig. 2C). Architecture and size both make a was blended with protonated linear polystyrene
ited solubility in organic solvents (13) on the clear difference in the miscibility of this system, of similar molecular weight and various concen-
order of 5 to 10 mg/ml. Solvent evaporation because the dendritic polyethylene nanoparticles trations of protonated polystyrene nanoparticles
from a fullerene-polymer solution will lead to a are miscible with 393-kD linear polystyrene (Fig. 3A). The radius of gyration for the linear
fullerene supersaturated state at low overall Eradius of gyration (Rg) 0 17.3 nm^, as evidenced polystyrene increases with nanoparticle concen-
concentration and likely phase separation by the nonfractal SANS results at small wave tration, and the linear chains remain globular in
(Fig. 1B). Thus, to reach the thermodynami- vector (16, 17). However, miscibility does not nature, as determined through careful analysis
cally favored state, we had to carefully control occur when the same polyethylene nanoparticles using models that distinguish between sphere-
the processing procedure for nanoparticle dis- are blended with either 155-kD (deuterated) lin- like, rod-like, and disk-like shapes (23).
persion to avoid a kinetically trapped condition. ear polystyrene (Rg 0 10.5 nm) or a smaller Chain stretching has been observed in some
A more surprising result is that we observed protonated 75-kD polystyrene (Rg 0 7.5 nm), as Monte Carlo simulations (24), although in many
dispersion of large branched polyethylene shown (Fig. 2C). The latter experiment demon- others chain contraction has been noted (25).
nanoparticles (14, 15) in polystyrene (Fig. 2 strates that the isotope effect (18) is not causing Both chain expansion and chain contraction have
and SOM). This is surprising because linear phase separation; rather, the relative size of the been observed in neutron scattering from iso-
polystyrene–linear polyethylene blends (2) nanoparticle and polymer is key (19–21). topically labeled polydimethylsiloxane (PDMS)
blends containing silica particles (26). In this sys-
tem, nanoparticles of radius È 1 nm were blended
A B with linear polymers with a similar Rg value of
È3 to 10 nm. The polymer mixtures with small-
er Rg values experienced chain contraction upon
nanoparticle addition, whereas the polymer mix-
tures with larger Rg values experienced chain
expansion with nanoparticle addition. In our sys-
tem, the linear polymer Rg value was varied
between 4 and 11 nm, the nanoparticle radius
was about 3 nm, and we observed chain expan-
100 nm 200 nm sion in all cases (Fig. 3A). Moreover, excluded
volume does not fully account for the radius in-
3
crease, because, if it is assumed that the indi-
10
vidual polymer and nanoparticle densities do not
Fig. 2. (A) Cartoon showing branched, dendritic C change on mixing, then the radius of gyration
polyethylene. (B) TEM of a 4 wt % blend of relative to that without nanoparticle incorpo-
2
dendritic polyethylene with 393-kD linear poly- 10 393 kDa polystyrene ration (Rg /Rg0) is expected to vary as (1 þ f) / .
1
3

styrene shows the individual polyethylene macro- The chain stretching is larger than the amount
I(q) (c m )

molecules with a size on the order of 20 to 30 nm.

-1

suggested by this relationship and is empirically

The Rg for the linear polystyrene is 17.4 nm and 10
1
Power law close to 1 þ cf, with c about 1.
so is larger than the dendritic polyethylene. 3.85 ± 0.02
Despite the linear polymer distortion, we
(Inset) A higher magnification. (C) Mixing with a
smaller molecular mass polystyrene (75 kD, Rg 0 0
found that a large miscible region exists for
10 cross-linked polystyrene nanoparticles blended
7.5 nm) produces phase separation. Power law
scattering of intensity (I) versus wave vector (q) is 75 kDa polystyrene with linear polystyrene (Fig. 3B). The data were
present at small wave vector for the lower mass determined from SANS through presence or ab-
-1
polystyrene, whereas the higher mass system 10 4 5 6 7 8 9 2 3 4 5 6 sence of fractal-like scattering at small wave vec-
-2
demonstrates miscibility without a power law 10 tor and at nanoparticle concentrations of 2 wt %.
q (Å )
-1
region. The intensity profile can be fitted with a Fractal-like behavior is indicative of a non-
polydisperse sphere model, yielding a mean radius (11.0 nm) for the dendritic polyethylene that equilibrium state consisting of irregular phase-
agrees well with the TEM images. separated aggregates, which exist on many length

www.sciencemag.org SCIENCE VOL 311 24 MARCH 2006 1741

 REPORTS
scales, despite the fact that the phase separation styrene blends, indicating that the negative mixing nanoparticles do not experience this enthalpic
is driven by a gain in equilibrium free energy. enthalpy is not due to isotopic substitution. driving force because AC tends toward A in this
Data from the fullerene–linear polystyrene and Furthermore, the c parameter is found to follow limit, confirming that C60 fullerenes are misci-
polyethylene nanoparticle–linear polystyrene Èþ0.01 – 6/T (in K), confirming that favorable ble solely through a favorable mixing entropy.
systems are also included in this phase diagram, enthalpic interactions, which are geometric in The mixing enthalpy of a nanoparticle, s,
indicating that this graph provides a useful guide origin, are responsible for the phase stability. can be related to the Flory parameter via s 0
for a range of nanocomposite systems. The ex- This unusual behavior is explained by con- ACc þ AU (c j enp/kBT ) K s0 – s1. Here s0 0
perimental data demonstrate that if the linear sidering the number of molecular contacts be- Ac is the insertion enthalpy in the absence of
polymer Rg is larger than the nanoparticle radius tween monomers in the isolated nanoparticle geometric effects due to uncovered area, and s1
then miscibility is promoted. Note that both the state compared to that of a nanoparticle in 0 AU
enp/kBT is the reduction in enthalpy
polymer Rg and nanoparticle radius were exper- the dispersed state, as illustrated in the ide- within the pure nanoparticle phase due to
imentally determined via SANS. alized model (Fig. 3C). The energy gain of a uncovered area. The areas are expressed in lat-
To experimentally determine the Flory c monomer-monomer contact is taken to be enp. tice site dimensions and so are dimensionless,
parameter, we found the second virial coefficient However, the van der Waals force operates over with s0 representing the insertion energy per
for 211-kD tightly cross-linked polystyrene an effective distance (d) so that only a fraction nanoparticle and c that of a monomer unit. Al-
nanoparticles dissolved in 473-kD deuterated of the nanoparticle area (AC ) 0 (zd/4a)
A has though the uncovered area can be quite small on
linear polystyrene to be 5.3
10j5 T 3.4
10j5 this favorable molecular contact. Here, z is the a given nanoparticle, the enthalpy gain via dis-
cc-mol/g2 at 127-C and 2.4
10j5 T 0.6
10j5 average coordination number of the nanoparti- persion can be substantial because of their large
cc-mol/g2 at 170-C by using SANS data and a cle aggregate. The remaining uncovered surface numbers, given by f/vnp with vnp being the nano-
Zimm analysis (16, 17). A standard analysis, area of the nanoparticle (AU K A – AC) does not particle volume (4/3pa3). So, the expected en-
strictly valid for linear-linear architecture blends profit from favorable contacts with other nano- thalpic stabilization is given by f/vnp
s1,
(27, 28), yields Flory parameters of c equal to particles. Nanoparticles may thus gain enthal- which exhibits a maximum at a nanoparticle
–2.7
10j3 (at 127-C) and –1.2
10j3 (at pically favorable monomer-monomer contacts radius of zd/2. This is truly a nanoscopic effect
170-C), demonstrating that mixing is favored by dispersion in the polymer melt, as has been because d is on the order of 1 nm and z is on the
at both temperatures. Bates and Wignall (18) noted in recent polymer reference interaction order of 6 for random packing (29), making the
found c to be È10j4 for deuterated poly- site model (PRISM) calculations (7). Smaller optimum radius for dispersion on the order of
3 nm, the size scale we have used in the present
study. If the nanoparticles are too small, then
21.0 kDa A B solubility suffers from too little or no uncovered
63.0 kDa
Polymer radius of gyration (nm)

1.15 15 tube area to achieve sufficient enthalpic gain, similar

155 kDa radius
to that experienced by the fullerenes, whereas a
B

Miscible
system containing larger particles has a reduced
1.10 mixing enthalpy by a reduction in nanoparticle
1+φ 10
Rg / Rg0

number for a given volume fraction.

The above argument hinges on uncovered
1.05
area developed by rigid particles (Fig. 3C) and is
5
1/3 Immiscible
related to an increase in the cohesive energy of a
[1+φ] material from its pure state. The dendritic poly-
1.00
ethylene nanoparticle system is a liquid at room
0
temperature, and so application of the cartoon to
0.00 0.02 0.04 0.06 0.08 0.10 0 5 10 15
this system is suspect. However, by using SANS
Volume fraction of nanoparticles Nanoparticle radius (nm)
of thermally annealed samples we measured the
C virial coefficient for this molecule dissolved in
Fig. 3. (A) The polymer radius of gyration (Rg) 393-kD linear polystyrene to be 2.1
10j5 T
relative to that without nanoparticles (Rg0) for three 1.7
10j5 cm3 mol/g2 at room temperature,
different molecular mass linear polystyrenes: 21, yielding c 0 –1.7
10j3 T 1.3
10j3. Again a
63, and 155 kD, as a function of volume fraction negative mixing enthalpy is found. This result is
(f) of 52.0-kD tightly cross-linked polystyrene rationalized first by noting the density of this
nanoparticles. The nanoparticles stretch the poly- material was determined to be quite small, 0.81 T
mer chains. The solid line represents the radius of 0.02 g/cm3, compared with the linear polyethyl-
gyration variation if the polymer density does not ene amorphous density extrapolated to room tem-
change upon mixing, and the behavior (1 þ f) / 1 3

perature (30), 0.86 to 0.89 g/cm3. Secondly, the

is expected. Instead, the data obeyed 1 þ cf, with melt surface tension (31, 32) at 160-C was mea-
c about 1. The error bars represent the error from a sured and found to be ,30% lower than that for
Guinier fit to the data at small wave vector. (B) A
polyethylene of similar molecular weight (mass
polymer radius of gyration–nanoparticle radius
of 236 kD). Both these results point to a reduced
phase diagram, with the solid circles representing
data where phase separation was detected and the internal energy in the isolated nanoparticle melt
open circles where miscibility occurs. Open circles by using the semi-empirical relation between the
with an
represent conditions where some agglomeration was detected by SANS yet large-scale phase surface tension (ST) and the cohesive energy
separation was not observed. Squares are the C60-polystyrene system; circles, the polystyrene nanoparticle– density (CED, internal energy per unit volume);
ST is about CED / (33) and is consistent with
2
3
polystyrene system; and triangles, the dendritic polyethylene–polystyrene system. The dashed line
represents the reptation tube radius, suggesting phase stability does not depend on the entanglement the enthalpy gain on mixing discussed above
structure. The nanoparticle fraction used to generate each data point was 2 wt %. (C) Cartoon (i.e., negative c).
illustrating that the attraction between pure nanoparticles is effective only over a fraction (AC) of the We generalized our observations by using a
available surface area (A) because of the limited range (d) over which dispersion forces operate. Flory lattice theory that has proven useful in

1742 24 MARCH 2006 VOL 311 SCIENCE www.sciencemag.org

 REPORTS
interpreting the phase morphology of complex References and Notes 28. T. P. Russell, R. S. Stein, J. Poly Sci. 20, 1593 (1982).
systems such as nanoparticle dispersion in poly- 1. P. J. Flory, Principles of Polymer Chemistry (Cornell Univ. 29. C. S. O’Hern, L. E. Silbert, A. J. Liu, S. R. Nagel,
Press, Ithaca, NY, 1953), p. 672. Phys. Rev. E 68, 011306 (2003).
mer blends, where reasonable agreement with 2. L. A. Utracki, Polymer Alloys and Blends (Hanser, 30. P. Zoller, D. J. Walsh, Standard Pressure-Volume-
lattice density functional theory has been dem- New York, 1990), p. 356. Temperature Data for Polymers (Technomic, Lancaster, PA,
onstrated (19). The first free energy term con- 3. L. H. Cragg, H. Hammerschlag, Chem. Rev. 39, 79 1995), p. 412.
sidered is the mixing entropy described by (1946). 31. G. T. Dee, B. B. Sauer, J. Colloid Inter. Sci. 152, 85
4. L. A. Utracki, B. Schlund, Polym. Eng. Sci. 27, 1512 (1992).
Fm/VkBT 0 Eflog(f)^/vvp þ E(1 – f)log(1 – f)^/N, (1987). 32. M. E. Mackay, G. Carmezini, B. B. Sauer, W. Kampert,
where Fm is the entropic gain due to mixing; N, 5. Y. Lin et al., Nature 434, 55 (2005). Langmuir 17, 1708 (2001).
the number of monomers in the linear poly- 6. J. B. Hooper, K. S. Schweizer, T. G. Desai, R. Koshy, 33. D. W. van Krevelen, Properties of Polymers, Their
mer; and V, the sample volume. Polymer chain P. Keblinski, J. Chem. Phys. 121, 6986 (2004). Estimation and Correlation with Chemical Structure
7. J. B. Hooper, K. S. Schweizer, Macromolecules 38, 8858 (Elsevier, New York, ed. 2, 1976), p. 620.
expansion (Fig. 3A) yields a loss of entropy, 34. G. Marrucci, Trans. Soc. Rheol. 16, 321 (1972).
(2005).
and so a Bstretching[ free energy cost (Fs ) for 8. N. S. Sariciftci, L. Smilowitz, A. J. Heeger, F. Wudl, 35. S. Y. Park, C. J. Barrett, M. F. Rubner, A. M. Mayes,
each molecule (34, 35) is included and approx- Science 258, 1474 (1992). Macromolecules 34, 3384 (2001).
imated by Fs/VkBT 0 3/2(1 – f)(Rg2/Rg02 – 1)/N. 9. N. S. Sariciftci, A. J. Heeger, Synth. Met. 70, 1349 (1995). 36. J. Huh, V. V. Ginzburg, A. C. Balazs, Macromolecules 33,
10. M. E. Mackay et al., Nat. Mater. 2, 762 (2003). 8085 (2000).
Previous work (21, 36) has used a chain stretch-
11. D. Weng et al., Eur. Polym. J. 35, 867 (1999). 37. W. C. K. Poon, J. Phys. Cond. Matter 14, R859 (2002).
ing term that has its basis in the analysis of 12. F. M. Du, J. E. Fischer, K. I. Winey, J. Poly Sci. Part B 41, 38. Supported by NSF DMR-0520415, NSF CTS-0400840,
polymer brushes, which does not show as large 3333 (2003). NSF NIRT-0210247, NSF-CTS-0417640, NSF
an Rg variation with f as we see experimentally. 13. V. N. Bezmel’nitsyn, A. V. Eletskii, Usp. Fiz. Nauk 41, NIRT-0506309, U.S. Department of Energy DE-FG02-
Combining the above terms, including the en- 1091 (1998). 90ER45418, DE-FG02-05ER46211, U.S. Department of
14. Z. Guan, P. M. Cotts, E. F. McCord, S. J. McLain, Science the Army, ARO W911NF-05-1-0357, NSF DMR-0135233
thalpy of mixing discussed above, we find that 283, 2059 (1999). (Z.G. CAREER), NSF CHEM-0456719, ARO DAAD19-01-1-
15. Z. B. Guan, J. Poly Sci. Part A 41, 3680 (2003). 0686, and the U.S. Department of Energy, Basic Energy
vnp F=VkB T 0 sfð1 j fÞ þ flogðfÞ þ 16. S. M. King, in Modern Techniques for Polymer Sciences–Materials Science, under contract W-31-
Characterisation, R. A. Pethrick, J. V. Dawkins, Eds. 109-ENG-38 to the University of Chicago. We thank
tð1 j fÞlogð1 j fÞ þ (Wiley, New York, 1999). P. Thiyagarajan, D. Wozniak, and developers of data
3 17. J. S. Higgins, H. C. Benoı̂t, Polymers and Neutron reduction routines at Argonne National Laboratory in
=2 tð1 j fÞðRg 2 =Rg0 2 j 1Þ ð1Þ Scattering (Clarendon Press, Oxford, 2002), p. 436. gathering and analyzing the neutron scattering data;
18. F. S. Bates, G. D. Wignall, Phys. Rev. Lett. 57, 1429 C. Popeney and D. Camacho for their help in synthesis
where t 0 vnp/N, which in terms of experimen- (1986). of the dendritic polyethylene; D. Bohnsack for its
19. V. Lauter-Pasyuk et al., Physica B (Amsterdam) 248, 243 characterization; A. Pastor for help with TEM;
tally accessible parameters is t 0 (a/Rg0)3r/r0, A. Frischknecht for valuable input on this work;
(1998).
where r is the bulk density of the linear poly- 20. R. B. Thompson, V. V. Ginzburg, M. W. Matsen, K. Schweizer for illuminating discussions and suggesting
styrene melt, and r0 is the density associated A. C. Balazs, Science 292, 2469 (2001). that we should have a negative mixing enthalpy
with one chain in the bulk polystyrene melt (i.e., 21. V. V. Ginzburg, Macromolecules 38, 2362 (2005). (c parameter) to be consistent with the observed
22. E. Harth et al., J. Am. Chem. Soc. 124, 8653 (2002). dispersion; and E. McGarrity for help with the figures.
the density of a single chain based on its Rg and
23. J. M. Gallas, K. C. Littrell, S. Seifert, G. W. Zajac,
molecular weight). The fact that t increases as P. Thiyagarajan, Biophys. J. 77, 1135 (1999). Supporting Online Material
the cube of the size ratio a/Rg0 implies that the 24. M. A. Sharaf, J. E. Mark, Polymer 45, 3943 (2004). www.sciencemag.org/cgi/content/full/311/5768/1740/DC1
stretching term is very unfavorable for a 9 Rg0, 25. M. Vacatello, Macromolecules 35, 8191 (2002). Materials and Methods
which is the basic reason why large nanopar- 26. A. I. Nakatani, W. Chen, R. G. Schmidt, G. V. Gordon, References
C. C. Han, Polymer 42, 3713 (2001).
ticles do not disperse, although this may not 27. W. A. Kruse, R. G. Kirste, J. Haas, B. J. Schmitt, D. J. Stein, 4 November 2005; accepted 18 January 2006
be generalized to colloidal systems where other Makrom. Chem. 177, 1145 (1976). 10.1126/science.1122225
physics comes into play (37).
For the case of polystyrene nanoparticles in
a linear polystyrene melt, we estimate the ef-
fect of the linear polymer stretch, the third term
in Eq. 1, by using the data of Fig. 3A approx-
Microheterogeneity of Singlet
imated by Rg/Rg0 , 1 þ cf. By setting the first
and second derivatives of Eq. 1 with respect Oxygen Distributions in Irradiated
to f to zero, one finds the binodal at fB 0
expA–E1 þ s þ (3c – 1)t^Z 0 expA–E1 þ s þ
(3c – 1)(r/r1)(a/Rg0)3^Z and the spinodal at
Humic Acid Solutions
fs 0 1/E2s þ (6c – 3c2 – 1)t^ for small f. In Douglas E. Latch and Kristopher McNeill*
Ginzburg_s work (21), an addition nonideal
term, the Carnahan-Starling term, is used to Singlet oxygen (1O2) is a highly reactive species formed through solar irradiation of organic
assess interaction effects between nanoparticles matter in environmental waters. Implicated in a range of reactions, it has proven difficult
that do not alter the binodal at small concentra- to quantify its spatial distribution in natural waters. We assessed the microheterogeneous
tion. Validity of the model is confirmed by distribution of 1O2 in irradiated solutions containing chromophoric dissolved organic matter
calculating s from the predicted binodal concen- (CDOM) by using molecular probes of varying hydrophobicity. The apparent 1O2 concentrations
tration, assuming c is 1 (Fig. 3A). Further, letting ([1O2]app), measured by recently developed hydrophobic trap-and-trigger chemiluminescent probe
a/Rg0 be 1 to delineate the phase boundary pffiffiffiffiffi molecules, were orders of magnitude higher than those measured by the conventional hydrophilic
(Fig. 3B) and calculating r=r0 ð0 1:7
M ) probe molecule furfuryl alcohol. The differential [1O2]app values measured by these probes reflect
for polystyrene (M is molecular mass in kD), then a steep concentration gradient between the CDOM macromolecules and the aqueous phase.
s values of –15 and –64 for molecular masses of A detailed kinetic model based on the data predicts probabilistic 1O2 distributions under different
30 kD (Rg0 0 5 nm) and 320 kD (Rg0 0 15 nm), solvent conditions.
respectively, are determined. Dividing these val-
ues by the nanoparticle area results in c on the inglet oxygen (1O2 ), or molecular oxygen cally produced in natural waters since the work
order of –3
10j3, in good agreement with the
values determined above via neutron scattering. S that is in its first electronic excited state
(1Dg ), has been known to be photochemi-
of Zepp et al. in 1977 (1). It is formed through
interaction of dissolved oxygen with excited

www.sciencemag.org SCIENCE VOL 311 24 MARCH 2006 1743