6874 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

Cite this: Chem. Soc. Rev., 2012, 41, 68746887

Colloidal nanoparticle clusters: functional materials by design

Zhenda Lu and Yadong Yin*

Received 1st June 2012

DOI: 10.1039/c2cs35197h

Signicant advances in colloidal synthesis made in the past two decades have enabled the

preparation of high quality nanoparticles with well-controlled sizes, shapes, and compositions.

It has recently been realized that such nanoparticles can be utilized as articial atoms for

building new materials which not only combine the size- and shape-dependent properties of

individual nanoparticles but also create new collective properties by taking advantage of their

electromagnetic interactions. The controlled clustering of nanoparticle building blocks into

dened geometric arrangements opens a new research area in materials science and as a result

much interest has been paid to the creation of secondary structures of nanoparticles, either by

direct solution growth or self-assembly methods. In this tutorial review, we introduce recently

developed strategies for the creation and surface modication of colloidal nanoparticle clusters,

demonstrate the new collective properties resulting from their secondary structures, and highlight

several of their many important technological applications ranging from photonics, separation,

and detection, to multimodal imaging, energy storage and transformation, and catalysis.

1. Introduction

Colloidal nanoparticles are of great interest for researchers

from a wide range of disciplines, including materials science,

chemistry, physics, and engineering, because of their unique

magnetic, electronic and optical properties, as compared to

their bulk counterparts. Signicant progress has been made in

the development of robust synthesis protocols which allow

precise control over composition, size, shape, surface proper-

ties, and uniformity of colloidal inorganic nanoparticles.1,2

Recently, the focus of synthetic eorts has been directed

towards the creation of secondary structures of colloidal

nanoparticles, which holds great promise for the development

of advanced materials with novel integrated functions.3

Clustering nanoparticles into secondary structures to form

so-called colloidal nanoparticle clusters (CNCs) not only

allows the combination of properties of individual nano-

particles but also takes advantage of the interactions between

neighboring nanoparticles which can result in new propertiesDepartment of Chemistry, University of California, Riverside CA92521, USA. E-mail: yadong.yin@ucr.edu

Zhenda Lu

Zhenda Lu received his BSand MS in Chemistry fromNanjing University in Chinain 2004 and 2007, respectively.He then came to the UnitedStates and is currently pursu-ing his PhD under the super-vision of Prof. Yadong Yin atUniversity of California,Riverside. His researchfocuses on the synthesis, sur-face modication and self-assembly of nanoparticles,and their bioanalytical andcatalytic applications. Yadong Yin

Yadong Yin received his BSand MS in Chemistry fromthe University of Science andTechnology of China in 1996and 1998, respectively, andthen PhD in Materials Scienceand Engineering from theUniversity of Washington in2002. He then worked as apostdoctoral fellow at theUniversity of California,Berkeley, and the LawrenceBerkeley National Labora-tory. In 2006 he joined theDepartment of Chemistry atUniversity of California,

Riverside. His research interests include colloidal chemistry,self-assembly, surface functionalization, and synthesis of nano-structured materials and their applications.

Chem Soc Rev Dynamic Article Links

www.rsc.org/csr TUTORIAL REVIEW

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online / Journal Homepage / Table of Contents for this issue

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6875

not present in the original constituents.4 A well-known example

is the assembly of noble metal nanoparticles into secondary

structures, which induces near eld coupling of surface plasmon

between adjacent particles. As a result, new optical properties

can be obtained, inducing shifts of plasmonic peaks and the

generation of hot spots that are very useful for enhancing

Raman scattering.57

Considering the fact that syntheses for a large variety of

nanoparticles have been developed in the past two decades and

the large number of dierent combinations that can be made

from these nanoparticles, one can easily see the great potential

of clustering approaches for the creation of novel nanoparticle-

based functional materials. Moreover, the forces involved in

nanoparticle clustering, including both covalent and non-covalent

interactions (e.g. hydrogen bonding, electrostatics and van der

Waals interactions) can be tailored by changing solvents,

surfactants, and reaction temperatures, providing exciting

opportunities for controlling specic geometric congurations

and consequently desired functions. Furthermore, the formation

of secondary structures may be able to eectively address many

challenges that are currently limiting the direct use of colloidal

nanoparticles in practical applications. For example, owing to

their high surface-to-volume ratio, small nanoparticles are widely

believed to possess signicantly enhanced catalytic activity. In

reality, however, the catalytic activity may quickly decay due to

the growth of nanoparticles as the result of interparticle sintering

during reactions. In addition, the capping ligands, which are

generally required to stabilize the nanoparticles during their

initial synthesis, may block access of the target molecules to

the catalyst surface and therefore severely reduce the catalytic

activity. We have recently shown that by organizing nanoparticle

catalysts into clusters we can circumvent these diculties by

allowing additional post-treatment to remove the capping

ligands, for example, by calcination at an appropriately high

temperature, while at the same time maintaining the high

surface area needed for high catalytic activity.8,9 Although

more eorts are still required to develop eective bottom-up

assembly approaches for colloidal nanoparticle clusters, this

strategy now opens up a nearly unlimited platform for designing

and manufacturing functional materials with new physical and

chemical properties.

This review will focus on the liquid-phase synthesis and

surface modication of various colloidal nanoparticle clusters,

which are typically composed of primary nanocrystallites of

approximately several to tens of nanometers in size. We also

highlight a number of representative examples of their many

potential technological applications, which may include catalysis,

energy storage and conversion, magnetic separation, multimode

imaging, chemical detection, and drug loading and release.

2. Synthesis of colloidal nanoparticle clusters

Colloidal nanoparticles are nanometer-scale inorganic nano-

particles, typically crystallites, stabilized by a layer of organic

capping ligands and dispersed in a solution. Pioneering work

on the synthesis of CdX (X = S, Se, Te) nanoparticles with

narrow size distributions in molten trioctylphosphine oxide

(TOPO) laid the foundation for the classic thermolytic routes,

which involve the reactions of inorganic precursors in organic

solvents at high temperatures.2 Many technologically important

high quality nanoparticles, such as semiconductor and metal oxide

nanocrystals, can now be routinely prepared through various

modied versions of the thermolytic method. Upon heating the

reaction solution to a suciently high temperature (typically

150320 1C), the precursors will be chemically transformed intoactive atomic or molecular species, which then condense to form

nanoparticles, the growth of which is strongly inuenced by the

presence of capping ligands. The size of nanoparticles can be

controlled by stopping the reaction at dierent growth stages or

changing the ligand concentrations. Shaped nanoparticles such as

nanodisks, nanorods, and nanoscale polyhedral structures can

also be synthesized by taking advantage of the selective adhesion

of certain ligands to particular crystalline facets to kinetically

control the relative growth rates along dierent crystalline

directions.1

The formation of secondary nanoparticle structures typically

involves more complex procedures or reaction pathways. As

shown in Fig. 1, there are basically two strategies for the

preparation of CNCs: (i) one-step processes which integrate the

synthesis of nanoparticles and their aggregation into clusters in a

single step; and (ii) multi-step processes which rst produce

nanoparticles with desired size, shape and surface functionality,

and then assemble them into clusters of designed congurations

in separate steps via methods such as solvent evaporation,

electrostatic attraction, or interfacial tension. While the one-step

processes are more ecient at producing CNC structures, the

multi-step processes have the advantage of being more exible

and universal for organizing nanoparticles of a large variety of

materials into CNCs with highly congurable structures.

2.1. Direct synthesis of nanoparticle clusters

Nanoparticle clusters can be produced through a number of

dierent one-step techniques, including thermolysis, solvothermal,

and microwave methods. Table 1 summarizes the various CNC

syntheses reported in the recent literature. Although the details in

these methods are dierent, every synthesis involves two growth

Fig. 1 Schematic illustration of the preparation strategies for colloidal

nanoparticle clusters (CNCs).

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

6876 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

stages in which primary nanoparticles rst nucleate and grow

in a supersaturated solution and then aggregate into larger

secondary particles.

2.1.1. Thermolysis method. A typical thermolysis process

entails reacting precursors in a solvent in the presence of a

surfactant at high temperatures. The reaction typically consists

of three critical components: precursors, organic capping

ligands, and solvents. The capping ligands bind to the nano-

particle surfaces, limit their growth, and prevent interparticle

agglomeration through steric interactions. With sucient

ligand protection, uniform nanoparticles, typically with dot

shapes, are obtained. However, by reducing the degree of ligand

protection to the domain of so-called limited ligand protec-

tion (LLP), complex three-dimensional (3D) nanostructures

can be produced through the oriented attachment of primary

nanoparticles. Peng et al. produced 3D nanoower-like struc-

tures for metal oxides such as In2O3, CoO, MnO and ZnO by

reducing the amount of stabilizing organic ligands to the point

that the primary nanoparticles were insuciently protected.10,11

Similar to the mainstream thermolytic syntheses, metal oxide

nanoparticles nucleate upon the thermolysis of precursors.

With increasing reaction temperature, the dot-shaped nano-

particles grow further at 250 1C and eventually agglomerateinto the ower-like clusters due to lack of sucient protection

from ligands. The key to CNC formation is to maintain an

appropriate concentration of capping ligands, which is not

enough to protect the primary nanoparticles against aggrega-

tion but sucient to stabilize the resulting 3D nanostructures.

The formation of relatively large crystalline clusters proceeds

through the 3D oriented attachment of primary nanoparticles.12

By changing the reaction conditions such as the specic concen-

tration of ligands and reaction time, the size of the clusters

can be adjusted within a reasonably wide range, for example,

1560 nm in the case of In2O3.10 The LLP approach is a

powerful strategy for the design of complex 3D CNCs, which

can be applied to metal oxides with dierent compositions. It is

also believed that the principle of LLP may be applicable to a

broad spectrum of colloidal nanoparticles, without involving

drastic alternations to the synthetic chemistry established for

simple 0D and 1D nanoparticles in the past decades.

Recently, our group has developed a one-pot high-tempera-

ture polyol process for the synthesis of polyelectrolyte-capped

superparamagnetic CNCs of magnetite (Fe3O4).13 Briey,

Fe3O4 CNCs were prepared by hydrolyzing FeCl3 with NaOH

atB220 1C in a diethylene glycol (DEG) solution with short-chain polyacrylic acid (PAA) as a surfactant. DEG was chosen

as the polar solvent because of its high boiling point as well as

its high permittivity, which enables high solubility for a variety

of polar inorganic and many organic compounds. Under the

reductive environment provided by DEG at a high tempera-

ture, Fe3+ partially transforms into Fe2+ and nally forms

Fe3O4 particles. The particle size can be tuned from 30 to

180 nm with a relatively narrow distribution by changing the

concentration of NaOH. The growth of CNCs follows the

well-documented two-stage growth model in which primary

nanoparticles nucleate rst in a supersaturated solution and

then aggregate into larger secondary particles. As shown in the

transmission electron microscopy (TEM) images in Fig. 2,

these magnetite CNCs have a well-developed cluster-like

structure: each cluster is composed of many interconnected

primary nanoparticles with a size of B10 nm. The crystallo-graphic alignment of the primary crystals relative to one another

has been observed in high resolution imaging and electron

diraction studies, suggesting that the possible formation

mechanism of CNCs involves oriented attachment and subsequent

high-temperature sintering during synthesis. This method has

been extended to the synthesis of CNCs of other materials,

such as PbS14 and ZnS.15

Recently, Kotovs group reported a one-step method for the

synthesis and self-assembly of monodispersed CdSe CNCs.16

Cadmium and selenium precursors were mixed at 80 1C in anaqueous solution with short and highly charged ligands such

as citrate anions, leading to the formation of nanoparticle

clusters with sizes tunable from 20 to 50 nm by changing

the reaction time. Similar to the two-stage growth model

for Fe3O4 CNCs discussed above, the assembly of CdSe

nanoparticle clusters occurs when primary nanoparticles are

present in the reaction media. It is important to note that the

polydispersity of the clusters (810%) was signicantly smaller

when compared with that of the primary nanoparticles of

which they are comprised (2530%). This self-limiting growth

Table 1 Summary of one-step approaches for CNC synthesis

Cluster Precursors Solvent and Surfactant Method T/1C Size range/nm Ref.

In2O3, ZnO, CoO, MnO2 Metal carboxylate(Ac, Mt or St)

ODE, OA or DA Thermolysis 250280 11

In2O3 Indium carboxylate(Ac, Mt or St)

ODE, DA Thermolysis 250290 1560 10

Fe3O4 FeCl3, NaOH DEG, PAA Thermolysis 220 30180 13PbS Pb(Ac)2, thiourea DEG, PAA Thermolysis 215 155240 14ZnO Zn(Ac)2, NaOH DEG, PAA Thermolysis 210 60180 15MFe2O4 (MQFe, Co, Mn, Zn) FeCl3, MCl2, NaAc EG, PEG Solvothermal 200 200800 23Fe3O4 Fe(acac)3 EG, PVP Solvothermal 140145 50100 18Fe3O4 FeCl3, NaAc EG, DEG, Sodium acrylate Solvothermal 200 6170 20MFe2O4 (MQFe, Mn, Zn, Co, Ni) FeCl3, MCl2, NaAc EG, DEG, PVP Solvothermal 200 20300 19Fe3O4 FeCl3, NaAc EG, Na3Cit Solvothermal 200 170300 21a-Fe2O3 FeCl3, urea THF, ethanol, PVP Solvothermal 180 22ZnO Zn(Ac)2 DEG Thermolysis (microwave) 57274 17

Abbreviations: Ac: acetate; Mt: myristate; St: stearate; ODE: 1-octadecene; OA: octadecyl alcohol; DA: decyl alcohol; EG: ethylene glycol; DEG:

diethylene glycol; PEG: polyethylene glycol; Cit: citrate; PAA: polyacyl acid; PVP: polyvinyl pyrrolidone.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6877

is enabled by a balance between electrostatic repulsion force

and van der Waals attraction force. This method has been

extended to the synthesis of CNCs of other materials, such as

CdS, ZnSe and PdS.

Thermolysis using microwave irradiation for heating represents

another method for producing colloidal inorganic nanomaterials,

which is very ecient and features unique advantages for

synthesis. It is believed that microwave dielectric heating can

address problematic issues such as heating inhomogeneity and

slow reaction kinetics present in conventional thermolysis

reactions, which rely on thermal conduction to drive chemical

reactions. As a result, it is becoming an increasingly popular

heating method for nanomaterial synthesis. Hu et al. have

employed a rapid microwave process to produce narrowly

distributed ZnO CNCs in relatively large quantities by heating

a zinc acetate solution in DEG using microwave irradiation.17

The high polarizability of DEGmakes this solvent an excellent

microwave absorbing agent, thus leading to a high heating rate

and short reaction time compared to existing solution-based

synthetic routes using conventional heating techniques. The size

of the clusters, which comprise small primary nanoparticles, can

be tuned continuously and precisely from about 57 to 274 nm

by simply varying the amount of zinc precursor.

2.1.2. Solvothermal synthesis. Solvothermal synthesis refers

to chemical reactions that are performed in a closed reaction

vessel (autoclave) at temperatures higher than the boiling point

of the solvent. This approach has become one of the most

widely used tools for nanoparticle synthesis due to the relatively

easy steps involved, simple setups, and reduced energy require-

ments, although it suers from several drawbacks such as

limited scalability and the lack of opportunities for direct

monitoring of the reaction process. A number of examples of

3D CNCs composed of primary nanoparticles have been

demonstrated through solvothermal methods.1823 In a typical

process for the synthesis of Fe3O4 CNCs, a solution containing

Fe(acac)3 (precursor), polyvinylpyrrolidone (PVP, surfactant),

and ethylene glycol (EG, solvent) was sealed in a Teon-lined

autoclave and heated to 140145 1C for 36 h.18 The products

were Fe3O4 CNCs containing disordered nanoscale pores

formed during assembly of the corresponding primary nano-

particles. Although the exact formation mechanism is unclear

due to the diculty of sampling under high temperature and

high pressure, it is convenient to control the size of the Fe3O4CNCs by changing the amount of precursor Fe(acac)3. To

further tune the sizes of the primary nanoparticles and secondary

Fe3O4 CNCs, Xuan et al. modied the solvothermal process

by utilizing sodium acrylate as surfactant to synthesize a series

of clusters.20 The average primary nanoparticle size could

be continuously tuned from B5.9 to B21.5 nm by simplychanging the weight ratio of sodium acrylate/NaAc, while the

overall size of the secondary structures could also be precisely

controlled in a wide range (up to B280 nm) by regulatingthe ratio of the two solvents (EG/DEG). Although this

solvothermal method is believed to be general for constructing

cluster structures from many other inorganic materials, the

success has been mainly limited to iron related materials, such

as Fe3O4, ferrite, and a-Fe2O3 (Table 1).

2.2. Clustering pre-synthesized nanoparticles

The utilization of pre-prepared nanoparticles as building

blocks for new materials such as 3D CNCs provides unique

opportunities to combine the inherent functionality of the

nanoparticles with potential collective properties resulting

from their interaction. Thanks to rapid progress in colloidal

nanostructure synthesis, a great number of materials can now

be routinely produced in the form of nanoparticles with excellent

control over size, shape and surface properties.1 It can easily be

appreciated that modular assembly approaches are highly attractive

for the preparation of secondary structured nanomaterials

with various congurations and programmable properties.

Many nanoparticle assembly methods have been developed

in the last decade. In this review, we focus on liquid-phase

strategies, which are very exible for controlling the structure,

composition and morphology of the nal CNC structures.

2.2.1. Evaporation-induced self-assembly (EISA). Self-assembly

of pre-synthesized nanoparticles through evaporation of solvents

in the presence of block-copolymers as structure directing

templates was initially designed for preparing mesoporous

metal oxide structures with high surface areas, high thermal

stability, and fully crystalline networks.24,25 Although meso-

porous materials have been extensively reported, the prepara-

tion of fully crystalline frameworks has remained a major

challenge due to the fact that many mesoporous oxide struc-

tures collapse during the crystallization process. Compared to

the well-known surfactant-templating method for mesoporous

silica structures, EISA utilizes crystalline nanoparticles instead

of molecular precursors as building blocks. In a typical

EISA process, monodisperse tin oxide nanoparticles of several

nanometers were prepared rst, and then dispersed in tetra-

hydrofuran (THF), forming a transparent and stable dispersion

with the addition of polybutadiene-block-poly(ethylene oxide)

(PB-PEO) block copolymer.26 The subsequent evaporation of

the THF solvent induced the assembly of nanoparticles and

PB-PEO block-copolymer micelles, nally leading to the formation

of mesoporous structures with ordered 1820 nm mesoscale pores.

The samples were further treated under air at high temperatures

Fig. 2 Representative TEM images of Fe3O4 CNCs with average

diameters of (a) 53, (b) 93, and (c) 141 nm. (d) High magnication

TEM images of 93 nm CNCs. Adapted with permission from ref. 13.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

6878 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

to completely remove the polymer templates, producing high

quality mesoporous structures.

EISA is a general method for the preparation of meso-

porous structures containing dierent nanoscale components.

When applied to nanoparticle dispersions conned in dened

volumes, such as droplets in an emulsion, it becomes a powerful

method to produce various mesoporous nanoparticle clusters

with desired overall dimensions. However, great eort is still

required to make these structures into a colloidal form with small

sizes and homogenous morphology in order to satisfy the needs

of specic applications, such as photonics and bioanalysis.

The oil-in-water emulsion evaporation method can be divided

into two steps as shown in Fig. 3a: (i) A nonpolar dispersion of

pre-synthesized nanoparticles is emulsied into an aqueous

solution containing emulsier (i.e., sodium dodecyl sulfate (SDS)

and cetyltrimethylammonium bromide (CTAB)), producing an

oil-in-water emulsion with oil droplets of a few micrometers.

(ii) The nanoparticles are concentrated and condensed into CNC

structures by evaporating the oil phase in the emulsion droplets.

The assembly is driven by the hydrophobic van der Waals

interactions of the capping ligands on the nanoparticle surface.

The hydrophobic nature of the nanoparticles also keeps the

clusters aggregated and prevents them from breaking up in the

aqueous environment. The emulsier is adsorbed onto the cluster

surface through the hydrophobichydrophobic interaction with

the capping ligands on the nanoparticles, which also helps to

disperse the clusters in water. Bai et al. demonstrated this facile

and universal bottom-up assembly strategy for preparing CNCs

from various nanoscale building blocks with dierent sizes and

shapes, such as BaCrO4, Ag2Se, CdS, PbS, Fe3O4, ZrO2, NaYF4nanodots, Bi2S3 and LaF3 nanoplates, and PbSeO3 nanorods.

27

The TEM images of BaCrO4 CNCs shown in Fig. 3b and c clearly

illustrate that the constituent nanoparticles retain their individual

character and do not sinter into larger units. The size of the CNCs

can be controlled by the parameters of the emulsication process

such as the concentration of nanoparticles in the oil phase and

the oil-to-water ratio. Specically, smaller clusters were

obtained by emulsication under sonication instead of stirring;

a higher nanoparticle concentration and oil-to-water ratio led

to larger clusters. More experimental details of this emulsion-

based nanoparticle assembly were studied by Simard and

co-workers.28 They concluded that: (1) the size and size

distribution of the clusters are dened by the droplets made

during emulsication and, as a result, are determined by the

emulsication conditions and emulsion composition; (2) the

size of the clusters is most conveniently controlled by varying

the concentration of nanoparticles in the oil phase; (3) the size

distribution can be narrowed by using a high volume fraction

of the droplet phase. This emulsion-based assembly process

also brings the convenience of incorporation of multiple compo-

nents into clusters to enable multifunctionality. Composite nano-

particle clusters can be fabricated by simply starting with amixture

of dierent types of nanoparticles, such as gFe2O3/TiO2,29

NaYF4-Yb,Er/NaYF4:Eu30 and CeO2/Pd.

31 The nanoparticle

packing characteristics in clusters can be determined by the

reaction temperature, which determines the rate of solvent

evaporation.32 Well-ordered nanoparticle superlattices with a

body-centered cubic (bcc) structure form with slow evapora-

tion at room temperature, while at a higher evaporation

temperature, multi-domain polycrystalline structures and

eventually completely amorphous structures will be produced.

To achieve more ordered packing of nanoparticles in clusters,

Cao and co-workers developed a modied assembly approach as

illustrated in Fig. 3c. Pre-synthesized uniform nanoparticles were

rst transferred from nonpolar solvent to aqueous solution by

using surfactants such as dodecyltrimethylammonium bromide

(DTAB). Then, ethylene glycol (EG) was added to the nano-

particle dispersion, leading to the formation of CNC structures

due to the weakened protection of DTAB in the EG solution.

Finally, the clusters were protected by adding PVP and annealed at

80 1C for 6 h.3,33,34 The annealing treatment is important forincreasing the order of the nanoparticle packing in the CNCs.

TEM images in Fig. 3d clearly show superlattice fringes, suggesting

that the nanoparticles were rearranged into a nearly perfect face-

centered cubic (fcc) packing after annealing.

Several modications to the emulsion evaporation method

were reported for fabricating CNC structures. Silica precursor

tetraethylorthosilicate (TEOS) was mixed with hydrophobic

nanoparticles in a nonpolar solvent and then emulsied in DEG

using a surfactant, as shown in Fig. 4a.35 This oil-in-DEG

technique oers the following advantages: (1) TEOS is constrained

Fig. 3 (a) Schematic illustration of oil-in-water emulsion solvent

evaporation for CNC preparation. (b) TEM images of BaCrO4 CNCs.

Adapted with permission from ref. 27. (c) Schematic illustration of CNC

preparation. (d) TEM images of CNCs 190 nm in diameter made of

Fe3O4 nanoparticles (5.8 0.2 nm in diameter) viewed along dierentzone axes. Scale bars: 20 nm. Adapted with permission from ref. 3.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6879

in the oil droplets together with the nanoparticles, ensuring

that the hydrolysis and condensation of TEOS only occurs in

the oil droplets, avoiding the formation of free silica spheres.

Silica was directly coated onto the CNCs after assembly.

(2) The use of DEG eectively limits the hydrolysis and

condensation of TEOS within the oil droplets, resulting in

better control of the particle size and avoidance of agglomeration.

In another case, biodegradable polymer (poly(D,L-lactic-co-glycolic

acid), PLGA) was introduced into the nanoparticle-containing oil

droplet to form an oil-in-water emulsion (Fig. 4b).36 After oil

evaporation, the nanoparticles were successfully embedded into

the PLGA matrix to form CNC structures. With the same

approach, QDs can be embedded in the matrix of polystyrene-

co-methacrylic acid (poly-St-co-MAA) copolymer.37 In these

assemblies, the polymers acted as glue for clustering nano-

particles and provided a matrix for loading drugs or other

functional species, such as uorescent probes. In addition to

serving as matrices, polymers can also be used as templates for

nanoparticle clustering.32 When polymers that were incompatible

with the nanoparticles were included in the emulsion formulation,

monolayer- and multilayer-nanoparticle coated polymer beads

and partially coated Janus beads were prepared. The nanoparticles

were expelled by the polymer as its concentration increased

upon evaporation of the solvent and accumulated on the

surfaces of the polymer beads (Fig. 4c). The number of

nanoparticle layers depended on the polymer/nanoparticle

ratio in the oil droplet phase.

2.2.2. Layer-by-layer (LBL) assembly. The layer-by-layer

(LBL) assembly technique was originally used for producing

thin polyelectrolyte lms on solid surfaces. The assembly

process involves sequential incubation of a charged solid

support in an oppositely charged polyelectrolyte solution.

After its invention, the LBL process was quickly adopted as

a versatile route for the creation of various nanoparticle shells

by sequential adsorption of nanoparticles and polyelectrolyte

onto the surface of submicrometer beads.3841 In a typical

process, submicrometer beads (e.g. silica or polystyrene) are rst

primed with several layers of polyelectrolyte lm to provide a

uniform charged surface that assists in the subsequent uniform

deposition of nanoparticles. Following nanoparticle adsorption, the

beads are centrifuged and washed for several cycles to remove

unadsorbed species, and then used for the next cycle of adsorption

of polyelectrolytes. The process is repeated until the desired number

of layers is obtained. An apparent limitation of the LBL assembly

method is that it typically only works with hydrophilic nano-

particles because it relies heavily on electrostatic interactions.

Many technologically important high quality nanoparticles,

Fig. 4 Schematic illustration of three emulsion evaporation based preparation methods for CNC structures and the corresponding TEM images

of the products: (a) TEOS assisted clustering, adapted with permission from ref. 35; (b) polymer assisted clustering, adapted with permission from

ref. 36; (c) polymer assisted clustering followed by phase segregation, adapted with permission from ref. 32.Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

6880 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

especially semiconductors and metal oxides, cannot be directly

assembled using the LBL method because they are predominantly

prepared and dispersed in organic solvents.

We recently developed a general LBL process that allows

convenient production of multifunctional composite particles by

direct self-assembly of hydrophobic nanoparticles on mercapto-

silica hosts containing high-density surface thiol groups.4 As

shown in Fig. 5a, hydrophobic nanoparticles can be directly

assembled onto the host surface through the strong coordination

interactions between soft metal cations and thiol groups. By

alternating mercapto-silica coatings and the nanoparticle

immobilization processes, multilayer structures composed of

various nanoparticles can be achieved. As a demonstration, we

started with 300 nm MPS spheres (mercapto-silica beads,

produced by hydrolysis and condensation of (3-mercaptopropyl)-

trimethoxysilane), immobilized g-Fe2O3 nanoparticles onthe surface, overcoated with a thin SiO2/MPS layer, and the

immobilized QDs on the surface. Fig. 5b and c show the electron

dispersive X-ray (EDX) elemental mapping and a typical TEM

image of a multilayer MPS@g-Fe2O3@SiO2&MPS@QD struc-ture. As indicated by the two dotted lines in Fig. 5c, the dierent

locations of Fe and Cd clearly suggest that the g-Fe2O3 nano-particles and QDs are distributed within dierent layers of the

composite. The gap between these two nanoparticle layers is 50 nm,

which corresponds to the thickness of the SiO2&MPS layer.

Yoon et al. recently developed a novel LBL method for

nanoparticle assembly based on the nucleophilic substitution reac-

tion between bromo and amine groups in organic media.42 They

rst prepared 2-bromo-2-methylpropionic acid (BMPA) stabilized

nanoparticles and amine-functionalized poly(amidoamine)

(PAMA) dendrimers, which were then sequentially coated

on colloidal silica beads. Analogous to the assembly induced

by the metalthiol interaction, the direct adsorption of the

nanoparticles in organic nonpolar solvent signicantly increases

their packing density in the lateral dimensions because electro-

static repulsion between neighboring nanoparticles is absent.

Moreover, the capping ligands on the nanoparticles are not

disturbed so that they retain their original properties such as

highly ecient luminescence.

2.2.3. Liquidliquid interface assembly. The assembly of

nanoparticles at a liquidliquid interface, analogous to the case

of Pickering emulsions, generates a resistant lm at the interface

between two immiscible phases, inhibiting the coalescence of

emulsion drops, as shown in Fig. 6a. A typical example is CdSe

nanoparticle assembly at the watertoluene interface to form a

kinetically stabilized water-in-oil emulsion (Fig. 6b and c).43 This

interfacial assembly is driven by the reduction in interfacial

energy, which depends on the nanoparticle size, particleparticle

interaction, particlewater and particleoil interactions. Larger

nanoparticles have a stronger stabilization eect for the assembly.

For example, 4.6 nm CdSe nanoparticles can be assembled on the

surface of an already stabilized droplet, displacing smaller 2.8 nm

particles.44 To fabricate mechanically stable capsules and

membranes from spherical nanoparticle assemblies, the nano-

particles need to be crosslinked at the interface, which requires

pre-modication of the nanoparticle surface with reactive

organic molecules.45 Compared with layer-by-layer polyelectro-

lyte deposition, assembly at the liquidliquid interface requires

fewer steps, aords ultrathin nanoparticle shells, and may reduce

structural defects due to the mobility of nanoparticles at the uid

interface. However, the emulsion droplets produced by liquid

liquid interface assembly are generally larger than several micro-

meters, which may limit their potential applications.

2.3. Comparison of one-step and multi-step methods

One-step syntheses of CNC structures are more convenient

and time-saving than those involving multiple steps. More-

over, clusters obtained through this method typically have

Fig. 5 (a) Schematic illustration showing the procedure of layer-by-

layer assembly of hydrophobic nanoparticles on MPS spheres.

(b and c) TEM and EDX mapping of the elemental distribution of

MPS@g-Fe2O3@SiO2&MPS@CdSe multilayer composites. Adaptedwith permission from ref. 4.

Fig. 6 (a) Schematic of the self-assembly of solid nanoparticles at the

oilwater interface. (b) Fluorescence confocal microscope image of

water droplets dispersed in toluene, covered with CdSe nanoparticles.

(c) Dierential interference contrast optical microscopy image of dried

droplets on a silicon substrate. Inset: AFM height section analysis.

Adapted with permission from ref. 43.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6881

narrower size distribution, which is very important for applica-

tions that require high uniformity, for example, in the construc-

tion of photonic crystals. However, the literature only contains a

limited number of examples for successful preparation of CNC

structures because controlling the clustering of the nanoparticles

during synthesis is usually even more challenging than that of

simple isolated nanoparticles. The key issue is to identify the

critical point of ligand protection for cluster formation. Above

the critical point, isolated and non-agglomerated nanoparticles

will form, whereas below the critical point, care has to be

taken to avoid the formation of uncontrolled aggregations

with random morphologies. Nanoparticle assembly through

two or more consecutive steps represents a more general class of

strategies for the preparation of CNC structures. Considering

the variety of nanoparticles that has been prepared in the last

decade, and the many possibilities to arrange them, such

modular approaches are advantageous for preparing materials

with tailored properties. Among various assembly approaches,

EISA in combination with emulsions represents a general

method which is very exible in organizing various nano-

particles into cluster structures, although it remains a challenge

to improve the uniformity of the resulting clusters. Assembly of

nanoparticles at a liquidliquid interface is unique in that it

produces hollow shells which can be further stabilized by

crosslinking the surface ligands. The challenge is in controlling

the size as well as the thickness of the shells. LBL assembly also

provides a universal strategy to arrange nanoparticles into

clusters with uniform size and morphology, but the loading

density of the nanoparticles is relatively low due to single layer

adsorption.

3. Surface modication of CNCs

Directly synthesized CNC structures are usually mechanically

stable and can be processed in the same manner as typical

colloidal nanoparticles, including multiple cleaning steps,

surface modication, and further assembly into more complex

structures. On the other hand, CNCs produced through

assembly approaches are generally protected by a layer of

surfactants that renders the particles highly dispersible in

solvents. The van der Waals interactions between the ligands

capping the nanoparticles and the hydrophobic tails of the

surfactant are generally weak and can be easily disturbed by

changes in the chemical environment, sometimes leading to

aggregation of the clusters in solution. The cluster structure

may be destroyed when subjected to strong mechanical forces

or when exposed to good solvents which can solvate individual

nanoparticles. In addition, it is often necessary to link func-

tional molecules to the surface of the CNCs, which is dicult

due to the weakly adsorbed surfactants. To address these

issues, a more robust protecting layer is often required. For

some applications, in particular, catalysis, a clean surface is

essential. Although calcination at high temperatures allows the

removal of the surfactants/capping ligands, it can lead to the

production of big aggregates. In this case, it may become

necessary to introduce a sacricial coating onto the surface of

the clusters which can prevent the formation of large aggrega-

tions. Here we summarize the three main surface treatment

methods reported for nanoparticle cluster structures: direct

calcination, silica coating, and polymer coating, as schemati-

cally illustrated in Fig. 7a.

3.1. Direct calcination

The application of nanostructured materials in bio-separation

or catalysis generally requires a clean surface to ensure sucient

active surface sites. However, high quality nanoparticles, as well

as CNCs assembled from them, are typically covered with a

layer of capping ligands, which prevents them from eectively

accepting target molecules. Direct calcination is the most

straightforward treatment to remove these organic ligands

and completely clean the material surface. Han et al. calcined

iron oxide clusters at 550 1C in air for 3 h to yield mesoporousmicrospheres with clean surfaces. The calcination removes

ligands occupying the materials surface and enhances the

mechanical stability of the clusters by bridging neighboring

nanoparticles through thermal fusion. On the other hand,

the primary nanoparticles still can be distinguished by TEM

imaging, suggesting that the interparticle fusion is modest.46

However, as expected, the calcination can cause severe aggrega-

tion of the clusters, and their spherical morphology may not be

well maintained during calcination.29

3.2. Silica coating

The usefulness of silica as a coating material mainly lies in its

high stability, easy control during the coating process,

chemical inertness, controllable porosity, processability and

optical transparency. In addition, a silica coating can endow a

composite with biocompatibility and the possibility of subsequent

functionalization. We have demonstrated a silica coating on

hydrophilic Fe3O4 CNCs by hydrolyzing tetraethoxysilane

(TEOS) in a mixture containing ethanol, CNCs, and ammonia

(NH3H2O) aqueous solution (Fig. 7b). The thickness of the silicashell can be tuned from ten to several hundred nanometers by

simply controlling the concentration of the precursor, TEOS.47,48

After silica coating, the CNCs can be well-dispersed in polar

solvents such as water and alcohol.

As an additional advantage, a silica shell provides more possibi-

lities for further surface modication through well-developed silane

chemistry. For example, we have demonstrated that a monolayer

of hydrophobic alkyl chains of n-octadecyltrimethoxysilane

Fig. 7 (a) Schematic illustration of surface treatment methods for

CNC structures. (b) TEM image of silica coated Fe3O4 CNCs.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

6882 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

(ODTMS) can be grafted onto the silica surface through

covalent SiOSi bonds, making the Fe3O4 CNC@SiO2colloids dispersible in most nonpolar solvents such as

1,2-dichlorobenzene, toluene, chloroform, and hexane.49 In

another demonstration, we have functionalized the Fe3O4CNC@SiO2 colloids with [3-(methacryloyloxy)propyl] trimethoxy-

silane (MPTMS) through siloxane linkage. An aqueous phase

precipitation polymerization process was then used to form a robust

thermoresponsive polymer coating on the core surface by

copolymerizing the surface MPTMS with N-isopropylacryl-

amide (NIPAM, monomer).50,51

A mesoporous silica shell can also be coated onto the CNCs

through a well-known surfactant-templating approach with

CTAB as the templating surfactant.52 An ordered mesoporous

silica phase with cylindrical channels is formed in the outer

layer, as conrmed by TEM imaging. These unique meso-

porous channels, which are perpendicular to the CNC core

surface, oer high surface area for the derivatization of

various functional groups, provide a large pore volume

for the adsorption and encapsulation of biomacromolecules

and even functional nanoparticles, and also enhance the

accessibility of the CNC cores.

For CNCs assembled from preformed nanoparticles, a silica

layer is critically important for maintaining their morphology

during calcination. We have recently developed a protected

calcination method to obtain readily dispersible colloidal

clusters by using hydrophobic TiO2 CNCs as a model system.8

In a typical process, TiO2 CNCs are prepared by evaporation

of the nonpolar solvent from an oil-in-water emulsion, and

then coated with a silica layer, calcined at 500 1C for 2 h in air,followed by removal of the SiO2 layer through chemical

etching in a dilute aqueous solution of NaOH. The silica

coating and removal steps are essential for the successful

fabrication of well-dispersible clusters. First, the silica layer

protects the clusters from aggregation during calcination at

high temperatures. Even though slight inter-cluster aggregation

occurs due to silica fusion during calcination, the subsequent

etching by NaOH removes the silica layer and releases the

clusters from aggregation. Second, the etching process after

calcination introduces a relatively high density of hydroxyl

groups so that the cluster surface becomes negatively charged,

making the clusters dispersible in water. This silica coating

calcination-silica removal method can be easily extended to

other clusters with dierent components or prepared by dierent

methods.

3.3. Polymer coating

Polymer coating is an alternative method to render clusters

more mechanically robust. In addition, polymer coating has a

number of other advantages: (1) the surface properties of

clusters can be easily tuned by coating with dierent polymer

layers. For example, polyethylene glycol (PEG) will greatly

enhance the water dispersity and biocompatibility of clusters.

(2) The large family of functional polymers oers many

opportunities for building up multifunctional clusters. (3) A

new functionality may also be incorporated into a polymer

shell by copolymerizing a functional monomer or through

post-modication methods. (4) The thickness of a polymer

shell can easily be adjusted down to several nanometers, which

is particularly important when only a thin shell is required. A

polymer coating can be achieved either by polymer adsorption

or monomer polymerization on the CNC surface. For example,

the positively charged poly(L-lysine)-poly(ethylene glycol)-folate

(PLL-PEG-FOL) can be adsorbed on negatively charged cluster

surfaces through the electrostatic interaction.36 An amphiphilic

hydrolyzed polymer, poly(maleic anhydride-alt-1-octadecene)

(PMAO), can be utilized to partially replace SDS coated

onto clusters through the coordination interaction between

carboxylic acid and metal oxide surfaces.53 Paquet et al.

reported a direct polymer coating for clusters using seed-

emulsion polymerization.28 In this method, CNCs were rst

prepared by the oil-in-water emulsion evaporation method

with SDS adsorbed on the surface. As the polymerization

reaction was thermally initiated, monomers such as methyl

methacrylate, styrene, and/or acrylic acid started to grow on

the cluster surface. To achieve polymerization at the surface of the

clusters and prevent nucleation and polymerization in micelles

formed by SDS, the concentration of the SDS in the dispersion of

clusters was maintained below the critical micelle concentration,

but high enough to maintain stability of the clusters.

When nanoparticles are originally covered with ligands

containing polymerizable groups, such as diynes and enediynes,

in situ photo-polymerization may be used to crosslink these

ligands after nanoparticle assembly.54,55 For example, Au

nanoparticles with protecting ligand 46-mercapto-22,43-

dioxo-3,6,9,12,15,18-hexaoxa-21,44-diazahexatetraconta-31,33-

diyn-1-oic acid (DA-PEG) were rst assembled into chain

structures by manipulating the electric dipoledipole inter-

actions, and then exposed to UV irradiation to crosslink the

surface ligands, thereby xing the cluster structure through the

polymerized DA-PEG thin layer and signicantly enhancing

their stabilities.54

4. Applications of CNCs

CNCs represent a new class of materials that have broad

applications in photonics, catalysis and bioanalysis due to

their unique properties compared to their primary nanoparticle

building blocks.

(1) CNCs can enhance the properties of the primary nano-

particles. For example, quantum dots (QDs) are attractive

uorescent materials for biological imaging due to their

spectral tunability in the visible and infrared regions. Individual

QDs, although possessing high quantum yields, sometimes are

insuciently bright due to their small sizes. However, CNC

structures assembled from primary QDs can provide much

stronger signals in biological imaging.56 Another important

case is superparamagnetic iron oxide nanoparticles, which have

primarily received attention for potential biomedical applica-

tions, as they are not subject to strong magnetic interactions in

dispersion. Several robust approaches have been developed for

synthesizing magnetic iron oxide (e.g., g-Fe2O3 or Fe3O4)nanoparticles with sizes ranging from several to B20 nm.However, these as-synthesized nanoparticles have a low

magnetization per particle, which limits their usage in many

important applications such as separation, targeted delivery or

magnetic resonance imaging (MRI). Increasing the nanoparticle

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6883

size increases the saturation magnetization, but also induces

the superparamagneticferromagnetic transition.13 Assembling

these nanoparticles into CNCs produces a material that possesses

a much higher saturated magnetization, but retains the original

superparamagnetic behavior of its building blocks even though

the overall cluster size exceeds 30 nm. By taking advantage of this

unique feature, many groups have successfully demonstrated the

use of these superparamagnetic iron oxide CNCs with various

sizes for magnetic separation.57,58 In addition, iron oxide CNCs

have shown improved contrast in MRI due to the high concen-

tration of nanoparticles in the cluster structures.59

(2) Clustering may enable multifunctionality by combining

various building blocks. For example, clusters composed of

magnetic iron oxide nanoparticles and uorescent quantum

dots have been widely studied as multiple-mode imaging

contrast agents for combining MRI and optical imaging.36

Replacing QDs with noble metal nanoparticles in such compo-

sites creates multifunctional structures that are capable of MRI

enhancement and photothermal therapy.60 Superparamagnetic

iron oxide nanoparticles were added to TiO2 clusters to facilitate

separation by an external magnetic eld for phosphopeptide

enrichment.29 Besides the building blocks of the clusters, capping

ligands on nanoparticles and the protection layer of the clusters

such as silica and polymer can also act as functional materials.

Clusters composed of oleic acid capped iron oxide nanoparticles

were employed for enrichment of peptides and proteins based

on the use of hydrophobichydrophobic interactions between

the oleic acid and the analytes.61 Aligned mesoporous silica

shells coated on paramagnetic clusters can be used for the

removal of microcystins.52

(3) CNCs may exhibit collective properties not present in

individual nanoparticles. A classic example is the clustering of

noble metal nanoparticles for generation of hot spots for

enhancing Raman scattering. The assembly of plasmonic

nanoparticles into secondary structures may induce near eld

electromagnetic coupling of surface plasmons between adjacent

particles, thus creating hot spots that can signicantly enhance

the Raman signals from analytes.62,63

(4) CNCs represent novel mesoporous structures with crystal-

line frameworks. Mesopores can be formed by packing primary

nanoparticles into clusters. For primary nanoparticles containing

capping ligands, they are typically calcined to remove the organic

ligands to allow full access by the target molecule. Calcination at

high temperatures may also enhance the mechanical stability of

the clusters by bridging neighboring nanoparticles together

through thermal fusion. Due to the crystalline nature of the

primary particles, they do not grow signicantly during calcina-

tion, preserving the high surface area and spherical morphology

of the clusters. The pore sizes of CNCs can be conveniently

controlled by changing the size and shape of the building blocks

during assembly. The submicrometer size of the clusters and the

three-dimensional pores enable fast diusion and adsorption of

target molecules. As a result of these great properties, clusters can

be employed for drug loading and delivery,64 bioseparation,8

sensing17 and catalysis.9 Furthermore, clustering methods can be

easily extended to the production of multicomponent structures

such as QD/TiO2 and QD/Au/TiO2 hybrid mesoporous

CNCs, which have been found to be highly ecient in photo-

electrochemical (PEC) cell applications.65,66

(5) Clusters facilitate surface modication. Ligand exchange

for individual nanoparticles usually involves several complex

steps and in many cases is detrimental to the physical proper-

ties of the nanoparticles because the new ligands may not be

able to eectively insulate the inorganic cores from chemical

disturbance from their environment. On the other hand,

surface modication of nanoparticle clusters can be considerably

easier as many approaches including ligand attachment, silica

encapsulation, and polymer coating have been well developed for

submicrometer objects.

These features have enabled a number of interesting appli-

cations for CNC structures in the last few years. Since it is

dicult to give a complete overview in this tutorial review,

here we use three typical applications to highlight their unique

advantages in designing structural and surface properties.

4.1. Magnetic responsive photonic crystal structures

The unique cluster structure allows Fe3O4 CNCs to retain

their superparamagnetism at room temperature even though

their overall size exceeds the critical size (30 nm) distinguishing

ferromagnetic and superparamagnetic magnetite. As shown in

Fig. 8a and b, the magnetization hysteresis loops of CNCs

with various sizes display typical superparamagnetic charac-

teristics with immeasurable remanence or coercivity at 300 K.

The cluster structure gives the Fe3O4 CNCs a much higher

saturated magnetization and thereby a stronger magnetic

response to external elds than the constituent nanoparticles.

The inset shows that the magnetic moment per cluster increases

with its overall size.

With the successful synthesis of Fe3O4 CNCs featuring the

superparamagnetic property, large and uniform sizes, and

highly charged surfaces, we have demonstrated their assembly

in aqueous solution into photonic crystal structures whose

optical signals can be instantly tuned by using external magnetic

elds.67,68 Under white light illumination, the colloidal photonic

crystals in the solution show brilliant colors from red to blue

when the strength of the applied magnetic eld is increased

(Fig. 8c). This visual eect, observable when viewed parallel to

the magnetic eld, results from the Bragg diraction of

incident light by the periodically ordered structures assembled

from Fe3O4 CNCs. A strong magnetic dipoledipole inter-

particle attraction is induced instantly in the superpara-

magnetic particle dispersion in response to the application of

external magnetic elds, which creates one-dimensional chains

each containing a string of particles (Fig. 8d). The interparticle

separation is dened by the balance between the magnetic

attraction and the interparticle repulsion of the electrostatic

force. By employing uniform superparamagnetic CNCs of appro-

priate sizes and surface charges, one-dimensional periodicity may

be created, which leads to strong diraction in the visible regime.

Magnetic forces, acting remotely over a large distance, not

only drive the rapid formation of colloidal photonic arrays

with a wide range of interparticle spacing, but also allow

instant tuning of the photonic properties by changing the

orientation of the colloidal assemblies or their periodicity

through the manipulation of the interparticle force balance.

Fig. 8e shows the reection spectra of 120 nm Fe3O4 CNC

aqueous solution in response to an external magnetic eld with

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

6884 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

varying strength achieved by changing the magnet-sample distance.

This optical response to the external magnetic eld is instantaneous

and fully reversible, and the required eld strength for realizing the

ordering of CNCs and color tuning is merely 50500 G. By

modifying the CNC surface property, we have been able to extend

this assembly process to solvents of various polarities,49 making it

possible to fabricate photonic crystal microspheres whose orienta-

tion and consequently photonic property can be easily controlled by

using external magnetic elds.69

4.2. Catalysis

Metal nanoparticles have been extensively studied as eective

catalysts in many reactions. In catalysis, it is important to ensure

that the dispersed metal nanoparticles retain their original struc-

ture, in particular their size and shape, throughout their pretreat-

ment, activation, and catalytic use. However, metal nanoparticles

tend to reconstruct, diuse, coalesce, and sinter during the reaction

process, which leads to signicant reduction in catalytic activity. It

is therefore highly desirable to develop ways to overcome

this limitation. Nanoparticle clusters are ideal support materials

because of their intrinsic porous structure, high surface area, rigid

framework, short diusion length for surrounding solutes, and

easy inclusion of catalyst nanoparticles. For example, metal

nanoparticles can be conveniently embedded in metal oxide

CNCs and display high and stable catalytic activity, as shown

in Fig. 9. The metal particles are eectively separated from

each other and trapped in the metal oxide matrix. Even after

heat treatment, the metal nanoparticles are still well separated.

In addition, the target molecules can easily access the metal

nanoparticle surface through the mesopores of the CNC

structures.31 The hydrogenation of cyclohexene to cyclohex-

ane and its dehydrogenation to benzene were used as probe

reactions to study the catalytic performance of the prepared

PdCeO2 composite CNCs. The results in Fig. 9 show excellent

selectivity of the CNC catalyst, with products being exclusively

cyclohexane at low reaction temperature (o185 1C) and benzeneat high temperature (B350 1C). Conducting the hydroconversionreactions for three cycles shows no signicant loss in catalytic

activity, indicating good thermal stability and robust performance

of the composite catalyst.

4.3. Bioseparation

Metal oxide anity chromatography (MOAC), built upon a

variety of metal oxide materials such as TiO2, has been

Fig. 8 (a) Schematic illustration showing that larger Fe3O4 CNCs have higher saturated magnetization. (b) Mass magnetization (M) as a function

of applied external eld (H) measured for 53 nm, 93 nm, 174 nm CNCs and a reference sample of 8 nm single crystalline nanoparticles of Fe3O4.

Inset shows the magnetic moment (m) per cluster (or particle) plotted in a logarithmic graph. Adapted with permission from ref. 13 (c).Photographs of aqueous solution of Fe3O4 CNCs in response to an increasing magnetic eld. The sample-magnet distance increases gradually from

left to right. (d) Schematic illustration showing the magnetic assembly of Fe3O4 CNCs into chains of periodically arranged particles which can

diract visible light. (e) Reection spectra of 120 nm Fe3O4 CNC aqueous solution in response to an external magnetic eld with varying strength

achieved by changing the magnet-sample distance. Adapted with permission from ref. 67.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6885

intensively studied because of its high selectivity for phospho-

peptide trapping. In addition, nanoparticles have many super-

ior characteristics for bioseparation compared to those of the

conventional micrometer-sized resins or beads, including high

capacity, fast and eective binding, and short diusion length

for biomolecules. However, there are several intrinsic dicul-

ties in the application of nanoparticles for bio-separation.

First, they cannot be conveniently separated from the solution

mixture by conventional methods such as centrifugation be-

cause of their extremely small sizes. Second, high quality

nanoparticles are typically synthesized in nonpolar solvents

so that they are covered with a layer of hydrophobic ligands,

which makes the particles non-water-soluble and greatly limits

their direct use in aqueous environments. Third, the surfaces

covered with hydrophobic ligands cannot eectively trap

biomolecules. Mesoporous CNCs with clean surfaces after

calcination can address these challenges. Using TiO2 CNCs

as an example, the high specicity and capacity of these

mesoporous TiO2 clusters have been demonstrated by eec-

tively enriching phosphopeptides from digests of phosphopro-

tein (a-casein), nonfat milk and human serum sample. Asshown in Fig. 10, after enrichment using the mesoporous TiO2clusters, phosphopeptides can be observed without any

obvious peaks from non-phosphopeptides, clearly showing

the eectiveness of the phosphopeptide enrichment.29 These

results conrm the excellent enrichment power of the nano-

particle clusters compared to solid TiO2 spheres, which can be

attributed to their high specic surface area and the clean TiO2surface. The outer surface of each cluster is made highly

hydrophilic to enhance the accessibility of the nanoparticle

clusters to phosphopeptides. The excellent performance of the

CNCs is also attributed to the submicron size of the clusters

and the three-dimensional pores which enable fast diusion

and adsorption of target molecules. By introducing a silica

coating/removal step, we were able to enhance the mechanical

stability of the clusters through calcination, and also make

their surface considerably charged to enable high water dis-

persibility. The calcination at high temperatures in air removes

the organic surfactants and makes the TiO2 surface fully

accessible to phosphopeptides. As a result, the porous TiO2clusters show attractive performance for selective enrichment

of phosphopeptides, with advantages including high adsorp-

tion capacity, high detection sensitivity, high selectivity, great

water dispersibility, high chemical/mechanical stability, and

easy separation from solution.

An inherent advantage of the self-assembly process is the

convenient incorporation of multiple components into the

clusters to further facilitate separation and detection. We have

also shown that the addition of superparamagnetic iron

oxide nanoparticles to the clusters allows not only selective

phosphopeptide enrichment but also their ecient removal

from the analyte solution by using an external magnetic eld.

Moreover, the pore sizes of the TiO2 clusters can be conve-

niently controlled by changing the size and shape of the

building blocks during assembly, and thus making it possible

to isolate the biomolecules such as intact phosphorylated

proteins with dierent sizes based on the size-exclusion

strategy.8

5. Conclusions and perspectives

We have reviewed the most recent strategies developed for the

preparation, surface modication, and application of colloidal

nanoparticle clusters. A number of liquid-phase synthesis

methods have been discussed, each of which has its own

advantages and drawbacks. While the one-step method is

straightforward and can produce uniform clusters, it is only

limited to a small group of materials. As a matter of fact,

the growth from nanoparticles to clusters is often more

challenging to control than that of the growth of individual

nanoparticles. In the multiple-step assembly strategy, the

synthesis and assembly of nanoparticles are carried out in

two or more consecutive steps. This exible preparation

strategy provides nearly limitless opportunities for producing

nanoparticle clusters from a large variety of materials and

their composites.

There are still many challenges that must be addressed

before CNCs reach their full potential in practical applica-

tions. The most critical problem is how to position specic

nanoparticles in desired locations within clusters, which is a

key for new collective properties resulting from nanoparticle

interactions. The second challenge is the development of

general methods that can produce uniform colloidal clusters

with controllable sizes and shapes. Some recent works have

clearly demonstrated the feasibility of assembling nanocrystals

into superstructures with dened shapes such as micron-sized

cubes, tetragonal structures,70,71 and dodecahedrons and

bipyramids.72 At the current stage, there is still room for

signicant improvement in controlling the size and uniformity

of CNCs. Similar to the development of synthesis methods

for individual colloidal nanoparticles, it is believed that

future research eorts may be directed to the production of

CNCs with particular shapes. Many interesting opportunities

may exist for shape-controlled CNCs, for example, for

constructing highly complex three-dimensional hierarchical

porous structures.

Fig. 9 Selectivity of the cyclohexene hydrogenation to cyclohexane

(lled symbols) and dehydrogenation to benzene (empty symbols) with

PdCeO2 CNCs as catalysts. Inset is a typical TEM image of hybrid

CNC structures. Adapted with permission from ref. 31.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

6886 Chem. Soc. Rev., 2012, 41, 68746887 This journal is c The Royal Society of Chemistry 2012

For clusters made using self-assembly approaches, another

challenge is the realization of ordering nanoparticles within

each cluster. The use of uniform primary nanoparticles is

apparently necessary, which however may not guarantee

perfect long range ordering. Many other parameters such as

solvents, temperature, and ligands may inuence the assembly

processes. Although very nice studies have been initiated on

this interesting concept,3,73 more eorts are still needed to

reveal the interactions involved during the assembly and how

they can be used to manipulate the superstructures built from

primary nanoparticles, which are often seen as articial

atoms in analogy to molecular building blocks. The structural

complexity may increase signicantly when nonspherical nano-

particles are used as building blocks for assembly. In addition, it

is expected that inclusion of impurity nanoparticles with

dierent sizes, shapes, compositions and surface properties

may induce the development of dierent polymorphic forms.

Eventually, it would be interesting to study the change in

physical properties associated with the crystal structural variation

in the formed clusters.

References

1 Y. Yin and A. P. Alivisatos, Nature, 2005, 437, 664.2 C. B. Murray, D. J. Norris andM. G. Bawendi, J. Am. Chem. Soc.,1993, 115, 8706.

3 J. Q. Zhuang, H. M. Wu, Y. G. Yang and Y. C. Cao, Angew.Chem., Int. Ed., 2008, 47, 2208.

4 Z. Lu, C. Gao, Q. Zhang, M. Chi, J. Y. Howe and Y. Yin, NanoLett., 2011, 11, 3404.

5 S. Lin, M. Li, E. Dujardin, C. Girard and S. Mann, Adv. Mater.,2005, 17, 2553.

6 S. Nie and S. R. Emory, Science, 1997, 275, 1102.7 P. Taladriz-Blanco, N. J. Buurma, L. Rodriguez-Lorenzo, J. Perez-Juste, L. M. Liz-Marzan and P. Herves, J. Mater. Chem., 2011,21, 16880.

Fig. 10 (a) Schematic illustration of the selective enrichment process of phosphorylated peptides by taking advantage of the special anity

between the phosphate groups and titania surface of the mesoporous CNCs. (b and c) Typical TEM images of TiO2 CNCs and solid spheres.

(df) MALDI mass spectra of the tryptic digest of a-casein (d) direct analysis, (e and f) analysis after enrichment using TiO2 CNCs and solidspheres. Adapted with permission from ref. 29.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online

This journal is c The Royal Society of Chemistry 2012 Chem. Soc. Rev., 2012, 41, 68746887 6887

8 Z. Lu, M. Ye, N. Li, W. Zhong and Y. Yin, Angew. Chem., Int.Ed., 2010, 49, 1862.

9 Q. Zhang, J.-B. Joo, Z. Lu, M. Dahl, D. Oliveira, M. Ye andY. Yin, Nano Res., 2011, 4, 103.

10 A. Narayanaswamy, H. F. Xu, N. Pradhan, M. Kim andX. G. Peng, J. Am. Chem. Soc., 2006, 128, 10310.

11 A. Naravanaswamy, H. Xu, N. Pradhan and X. Peng, Angew.Chem., Int. Ed., 2006, 45, 5361.

12 R. L. Penn and J. F. Baneld, Science, 1998, 281, 969.13 J. Ge, Y. Hu, M. Biasini, Ward P. Beyermann and Y. Yin, Angew.

Chem., Int. Ed., 2007, 46, 4342.14 C. Li, Y. Zhao, F. Li, Z. Shi and S. Feng,Chem.Mater., 2010, 22, 1901.15 C. Li, Y. Zhao, L. Wang, G. Li, Z. Shi and S. Feng, Eur. J. Inorg.

Chem., 2010, 217.16 Y. Xia, T. D. Nguyen, M. Yang, B. Lee, A. Santos, P. Podsiadlo,

Z. Tang, S. C. Glotzer and N. A. Kotov, Nat. Nanotechnol., 2012,7, 479.

17 X. L. Hu, J. M. Gong, L. Z. Zhang and J. C. Yu, Adv. Mater.,2008, 20, 4845.

18 Y. Zhu, W. Zhao, H. Chen and J. Shi, J. Phys. Chem. C, 2007,111, 5281.

19 S. H. Xuan, F. Wang, Y. X. J. Wang, J. C. Yu and K. C. F. Leung,J. Mater. Chem., 2010, 20, 5086.

20 S. Xuan, Y.-X. J. Wang, J. C. Yu and K. Cham-Fai Leung, Chem.Mater., 2009, 21, 5079.

21 J. Liu, Z. K. Sun, Y. H. Deng, Y. Zou, C. Y. Li, X. H. Guo,L. Q. Xiong, Y. Gao, F. Y. Li and D. Y. Zhao, Angew. Chem., Int.Ed., 2009, 48, 5875.

22 X.-L. Fang, C. Chen, M.-S. Jin, Q. Kuang, Z.-X. Xie, S.-Y. Xie,R.-B. Huang and L.-S. Zheng, J. Mater. Chem., 2009, 19, 6154.

23 H. Deng, X. Li, Q. Peng, X. Wang, J. Chen and Y. Li, Angew.Chem., Int. Ed., 2005, 44, 2782.

24 A. Corma, P. Atienzar, H. Garcia and J. Y. Chane-Ching, Nat.Mater., 2004, 3, 394.

25 A. S. Deshpande, N. Pinna, B. Smarsly, M. Antonietti andM. Niederberger, Small, 2005, 1, 313.

26 J. H. Ba, J. Polleux, M. Antonietti and M. Niederberger, Adv.Mater., 2005, 17, 2509.

27 F. Bai, D. S. Wang, Z. Y. Huo, W. Chen, L. P. Liu, X. Liang,C. Chen, X. Wang, Q. Peng and Y. D. Li, Angew. Chem., Int. Ed.,2007, 46, 6650.

28 C. Paquet, L. Page, A. Kell and B. Simard, Langmuir, 2009,26, 5388.

29 Z. Lu, J. Duan, L. He, Y. Hu and Y. Yin, Anal. Chem., 2010,82, 7249.

30 P. Li, Q. Peng and Y. Li, Adv. Mater., 2009, 21, 1945.31 C. Chen, C. Nan, D. Wang, Q. Su, H. Duan, X. Liu, L. Zhang,

D. Chu, W. Song, Q. Peng and Y. Li, Angew. Chem., Int. Ed., 2011,50, 3725.

32 T. Isojima, S. K. Suh, J. B. V. Sande and T. A. Hatton, Langmuir,2009, 25, 8292.

33 J. Q. Zhuang, H. M. Wu, Y. A. Yang and Y. C. Cao, J. Am. Chem.Soc., 2007, 129, 14166.

34 J. Zhuang, A. D. Shaller, J. Lynch, H. Wu, O. Chen, A. D. Q. Liand Y. C. Cao, J. Am. Chem. Soc., 2009, 131, 6084.

35 L. Li, E. S. G. Choo, J. Yi, J. Ding, X. Tang and J. Xue, Chem.Mater., 2008, 20, 6292.

36 J. Kim, J. E. Lee, S. H. Lee, J. H. Yu, J. H. Lee, T. G. Park andT. Hyeon, Adv. Mater., 2008, 20, 478.

37 Y. Zhao, Y. Ma, H. Li and L. Wang, Anal. Chem., 2012, 84, 386.38 N. A. Kotov, I. Dekany and J. H. Fendler, J. Phys. Chem., 1995,

99, 13065.39 D. E. Gomez, I. Pastoriza-Santos and P. Mulvaney, Small, 2005,

1, 238.

40 F. Caruso, M. Spasova, V. Salgueirino-Maceira and L. M. Liz-Marzan,Adv. Mater., 2001, 13, 1090.

41 F. Caruso, R. A. Caruso and H. Mohwald, Science, 1998,282, 1111.

42 M. Yoon, Y. Kim and J. Cho, ACS Nano, 2011, 5, 5417.43 Y. Lin, A. Boker, H. Ska, D. Cookson, A. D. Dinsmore,

T. Emrick and T. P. Russell, Langmuir, 2005, 21, 191.44 Y. Lin, H. Ska, T. Emrick, A. D. Dinsmore and T. P. Russell,

Science, 2003, 299, 226.45 H. Ska, Y. Lin, R. Tangirala, K. Breitenkamp, A. Boker,

T. P. Russell and T. Emrick, Adv. Mater., 2005, 17, 2082.46 L. Han, Z. Shan, D. H. Chen, X. J. Yu, P. Y. Yang, B. Tu and

D. Y. Zhao, J. Colloid Interface Sci., 2008, 318, 315.47 J. P. Ge and Y. D. Yin, Adv. Mater., 2008, 20, 3485.48 R. Fu, X. Jin, J. Liang, W. Zheng, J. Zhuang and W. Yang,

J. Mater. Chem., 2011, 21, 15352.49 J. P. Ge, L. He, J. Goebl and Y. D. Yin, J. Am. Chem. Soc., 2009,

131, 3484.50 B. Luo, X.-J. Song, F. Zhang, A. Xia, W.-L. Yang, J.-H. Hu and

C.-C. Wang, Langmuir, 2009, 26, 1674.51 J. Ge, T. Huynh, Y. Hu and Y. Yin, Nano Lett., 2008, 8, 931.52 Y. Deng, D. Qi, C. Deng, X. Zhang and D. Zhao, J. Am. Chem.

Soc., 2008, 130, 28.53 Q. Zhang, X. Wang and Y. Zhu, J. Mater. Chem., 2011, 21, 12132.54 R. Fernandes, M. Li, E. Dujardin, S. Mann and A. G. Kanaras,

Chem. Commun., 2010, 46, 7602.55 J. M. Walker, L. Gou, S. Bhattacharyya, S. E. Lindahl and

J. M. Zaleski, Chem. Mater., 2011, 23, 5275.56 X. Yan, Y. Cui, W. Qi, Y. Su, Y. Yang, Q. He and J. Li, Small,

2008, 4, 1687.57 M. M. Ye, Q. Zhang, Y. X. Hu, J. P. Ge, Z. D. Lu, L. He,

Z. L. Chen and Y. D. Yin, Chem.Eur. J., 2010, 16, 6243.58 J. P. Ge, Q. Zhang, T. R. Zhang and Y. D. Yin, Angew. Chem., Int.

Ed., 2008, 47, 8924.59 P. H. Qiu, C. Jensen, N. Charity, R. Towner and C. B. Mao,

J. Am. Chem. Soc., 2010, 132, 17724.60 J. Kim, J. E. Lee, J. Lee, Y. Jang, S. W. Kim, K. An, H. H. Yu and

T. Hyeon, Angew. Chem., Int. Ed., 2006, 45, 4789.61 Z. Lu, L. He and Y. Yin, Chem. Commun., 2010, 46, 6174.62 B. Yan, A. Thubagere, W. R. Premasiri, L. D. Ziegler, L. Dal

Negro and B. r. M. Reinhard, ACS Nano, 2009, 3, 1190.63 C. Yuan, R. Liu, S. Wang, G. Han, M.-Y. Han, C. Jiang and

Z. Zhang, J. Mater. Chem., 2011, 21, 16264.64 B. Luo, S. A. Xu, A. Luo, W. R. Wang, S. L. Wang, J. Guo,

Y. Lin, D. Y. Zhao and C. C. Wang, ACS Nano, 2011, 5, 1428.65 L. Liu, J. Hensel, R. C. Fitzmorris, Y. Li and J. Z. Zhang, J. Phys.

Chem. Lett., 2010, 1, 155.66 L. Liu, G. Wang, Y. Li, Y. Li and J. Z. Zhang, Nano Res., 2011,

4, 249.67 J. Ge, Y. Hu and Y. Yin, Angew. Chem., Int. Ed., 2007, 46, 7428.68 J. Ge, Y. Hu, T. Zhang, T. Huynh and Y. Yin, Langmuir, 2008,

24, 3671.69 J. P. Ge, H. Lee, L. He, J. Kim, Z. D. Lu, H. Kim, J. Goebl,

S. Kwon and Y. D. Yin, J. Am. Chem. Soc., 2009, 131, 15687.70 F. Dumestre, B. Chaudret, C. Amiens, P. Renaud and P. Fejes,

Science, 2004, 303, 821.71 S. Disch, E. Wetterskog, R. l. P. Hermann, G. Salazar-Alvarez,

P. Busch, T. Bruckel, L. Bergstrom and S. Kamali, Nano Lett.,2011, 11, 1651.

72 A. Pucci, M.-G. Willinger, F. Liu, X. Zeng, V. Rebuttini,G. Clavel, X. Bai, G. Ungar and N. Pinna, ACS Nano, 2012,6, 4382.

73 Y. Nagaoka, O. Chen, Z. Wang and Y. C. Cao, J. Am. Chem. Soc.,2012, 134, 2868.

Dow

nloa

ded

by U

nive

rsity

of C

alga

ry o

n 01

/05/

2013

09:

09:1

5.

Publ

ished

on

06 A

ugus

t 201

2 on

http

://pu

bs.rs

c.org

| doi:

10.103

9/C2C

S3519

7HView Article Online