713 Puriﬁcation of Carbon Nanotubes

8/12/2019 713 Purification of Carbon Nanotubes

1/23

Purification of carbon nanotubes

Peng-Xiang Hou, Chang Liu, Hui-Ming Cheng*

Shenyang National Laboratory for Materials Science, Institute of Metal Research, Chinese Academy of Sciences,

72 Wenhua Road, Shenyang 110016, PR China

A R T I C L E I N F O

Article history:

Received 27 June 2008Accepted 1 September 2008

Available online 9 September 2008

A B S T R A C T

It is predicted theoretically and understood experimentally that carbon nanotubes (CNTs)

possess excellent physical and chemical properties and have wide-range potential applica-

tions. However, only some of these properties and applications have been verified or real-

ized. To a great extent, this situation can be ascribed to the difficulties in getting high-

purity CNTs. Because as-prepared CNTs are usually accompanied by carbonaceous or

metallic impurities, purification is an essential issue to be addressed. Considerable pro-

gress in the purification of CNTs has been made and a number of purification methods

including chemical oxidation, physical separation, and combinations of chemical and

physical techniques have been developed for obtaining CNTs with desired purity. Here

we present an up-to-date overview on the purification of CNTs with focus on the principles,

the advantages and limitations of different processes. The effects of purification on the

structure of CNTs are discussed, and finally the main challenges and developing trends

on this subject are considered. This review aims to provide guidance and to stimulate inno-

vative thoughts on the purification of CNTs.

2008 Elsevier Ltd. All rights reserved.

Contents

1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2004

1.1. CNT synthesis techniques. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2004

1.2. Impurities coexisting with CNTs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2004

1.3. Assessment of CNT purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2005

1.4. Purpose of this review. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2006

2. Purification methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2007

2.1. Chemical oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2007

2.1.1. Gas phase oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2007

2.1.2. Liquid phase oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2009

2.1.3. Electrochemical oxidation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2011

2.1.4. Brief summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2011

2.2. Physical-based purification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2011

2.2.1. Filtration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2012

2.2.2. Centrifugation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2013

2.2.3. Solubilization of CNTs with functional groups. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2013

2.2.4. High temperature annealing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2013

0008-6223/$ - see front matter

2008 Elsevier Ltd. All rights reserved.doi:10.1016/j.carbon.2008.09.009

* Corresponding author: Fax: +86 24 2390 3126.E-mail address:[email protected](H.-M. Cheng).

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5

a v a i l a b l e a t w w w . s c i e n c e d i r e c t . c o m

j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / c a r b o n
mailto:[email protected]:[email protected]


2/23

2.2.5. Other physical techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2014

2.2.6. Combination of purification and separation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2014

2.2.7. Brief summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2015

2.3. Multi-step purification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2016

2.3.1. HIDE-assisted multi-step purification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2016

2.3.2. Microfiltration in combination with oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2016

2.3.3. Sonication in combination with oxidation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2017

2.3.4. High temperature annealing in combination with extraction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2018

2.3.5. Brief summary. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2018

2.4. Applicability of typical purification techniques . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2018

3. Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2020

3.1. Synthesis methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2020

3.2. Purification methods. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2021

3.3. Purity assessment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2021

4. Concluding remarks. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2021

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2021

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2021

1. Introduction

Elemental carbon in sp2 hybridization can form a variety of

amazing structures, such as graphite (3D), graphene (2D), car-

bon nanotubes (CNTs, 1D) and fullerene (0D). CNTs defined by

Iijima in 1991[1] have a unique tubular structure with nano-

meter scale diameters and large length/diameter ratios. CNTs

may consist of one (single-walled CNTs, SWCNTs) or up to

tens and hundreds (multi-walled CNTs, MWCNTs) seamless

graphene cylinders concentrically stacked with an adjacent

layer spacing of 0.34 nm. Owing to the covalent sp2 bonds

formed between individual carbon atoms, CNTs are stiffer

and stronger potentially than any other known materials.Thus, CNTs have ultra-high Youngs modulus and tensile

strength, which makes them promising in serving as a rein-

forcement of composite materials with desired mechanical

properties. Because of the symmetry and unique electronic

structure of graphene, the structure of a SWCNT determines

its electrical properties. For a SWCNTwith a given (n,m) index

[2], when (2n+m) = 3q(qis an integer), the nanotube is metal-

lic, otherwise the nanotube is a semiconductor. Not only do

these nanotubes show amazing mechanical and electronic

properties, but also possess well-defined hollow interiors

and biocompatibility with living systems. As a result, CNTs

are considered to be excellent candidates for many potential

applications, including but not limited to: catalyst and cata-lyst supports [3,4], composite materials [5,6], sensors and

actuators [7,8], field emitters [9,10], tips for scanning probe

microscopy[11,12], conductive films[13,14], bio-nanomateri-

als[15], energy storage media[16,17]and nanoelectronic de-

vices[18,19].

1.1. CNT synthesis techniques

Nowadays, CNTs can be produced in large quantities by three

dominant techniques: chemical vapor deposition (CVD,

including high-pressure carbon monoxide (HiPco) process)

[20], arc discharge [1], and laser ablation [21]. CVD involves

catalyst-assisted decomposition of hydrocarbons (commonly

benzene, ethanol, acetylene, propylene, methane, ethylene,

CO, etc.) and growth of CNTs over the catalyst (usually transi-

tion metals such as Ni, Fe, Co, etc.) in a temperature range of

3001200 C. Good alignment[22]as well as positional control

on a nanometric scale [23] can be achieved by using CVD.

Control over diameter, shell number, and growth rate of CNTs

are also realized with this method. The chief drawback of

CVD is the high defect density of the obtained CNTs owing

to low synthesis temperatures, compared with arc discharge

and laser ablation. As a result, the tensile strength of the

CNTs synthesized by CVD is only one-tenth of those made

by arc discharge[24]. Typical SWCNT content in as-prepared

samples by CVD is 3050 wt%, while the content ofMWCNTs is in the range of 3099 wt% depending on their

diameters. The by-products are usually aromatic carbon,

amorphous carbon, polyhedral carbon, metal particles, etc.

Arc discharge uses two electrodes (at least one electrode is

made of graphite) through which a direct current (DC) is

passed in a gaseous atmosphere. MWCNTs can be obtained

by arc discharge without any metal catalyst, while mixed me-

tal catalysts inserted into the anode are required when syn-

thesizing SWCNTs by this method. In laser ablation for

producing CNTs, an intense laser beam is used to ablate/

vaporize a target consisting of a mixture of graphite and me-

tal catalyst in a flow of inert gas. This method favors the

growth of SWCNTs with controlled diameter depending onreaction temperature[24]. When using arc discharge and laser

ablation for SWCNT synthesis, side products such as fuller-

enes, amorphous carbon, graphite particles, and graphitic

polyhedrons with enclosed metal particles are also formed.

The record high-purity of the SWCNTs synthesized by arc dis-

charge has been reported to be 80% by volume[25].

1.2. Impurities coexisting with CNTs

As-synthesized CNTs prepared by the above methods inevita-

bly contain carbonaceous impurities and metal catalyst parti-

cles, and the amount of the impurities commonly increases

with the decrease of CNT diameter. Carbonaceous impurities

2004 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


3/23

typically include amorphous carbon, fullerenes, and carbon

nanoparticles (CNPs) (as shown inFig. 1). Because the carbon

source in arc discharge and laser ablation comes from the

vaporization of graphite rods, some un-vaporized graphitic

particles that have fallen from the graphite rods often exist

as impurity in the final product. In addition, graphitic polyhe-

drons with enclosed metal particles also coexist with CNTs

synthesized by arc discharge and laser ablation as well as

high temperature (>1000 C) CVD. Fullerenes can be easily re-

moved owing to their solubility in certain organic solvents.

Amorphous carbon is also relatively easy to eliminate be-

cause of its high density of defects, which allow it to be oxi-

dized under gentle conditions. The most knotty problem is

how to remove polyhedral carbons and graphitic particles

that have a similar oxidation rate to CNTs, especially

SWCNTs. Metal impurities are usually residues from the

transition metal catalysts. These metal particles are some-

times encapsulated by carbon layers (varying from disordered

carbon layers to graphitic shells, as shown inFig. 1b and c)

making them impervious and unable to dissolve in acids. An-

other problem that needs to be overcome is that carbona-

ceous and metal impurities have very wide particle size

distributions and different amounts of defects or curvature

depending on synthesis conditions, which makes it rather dif-

ficult to develop a unified purification method to obtain repro-

ducibly high-purity CNT materials. To fulfill the vast potential

applications and to investigate the fundamental physical and

chemical properties of CNTs, highly efficient purification of

the as-prepared CNTs is, therefore, very important.

1.3. Assessment of CNT purity

To evaluate the purity of CNTs, the efficiency of a purification

method as well as changes in the structure of CNTs during

purification, characterization methods with rapid, convenient

and unambiguous features are urgently required. Character-

ization of CNT samples falls into three groups: metal catalyst,

carbonaceous impurity, and CNT structure variation (defects,

functional groups, cap opening, cutting, etc.). Their character-

ization mainly depends on electron microscopy (EM, including

scanning EM (SEM), and transmission EM (TEM)), thermogravi-

metric analysis (TGA), Raman spectroscopy and ultraviolet-

visible-near infrared (UVvis-NIR) spectroscopy.

EM is a useful technique allowing for direct observations of

impurities, local structures as well as CNT defects. Owing to

the small volume of sample analyzed and the absence of algo-

rithms to convert images into numerical data, EM cannot give

a quantitative evaluation of the purity of CNTs [28].

TGA is effective in evaluating quantitatively the quality of

CNTs, in particular, the content of metal impurity. It is easy

and straightforward to obtain the metal impurity content

using TGAby simplyburning CNTsamples in air. A higher oxi-

dation temperature (>500 C) is always associated with purer,

less defective CNT samples. The homogeneityof CNT samples

can be evaluated by standard deviations of the oxidation tem-

perature and metal content obtained in several separate TGA

runs [29]. The real difficulty is qualitative or quantitative

assessment of carbonaceous impurity, which is influenced

by the amount of defects, forms of carbon, and so on.

Raman spectroscopy is a fast, convenient and non-destruc-

tive analysis technique. To some extent, it can quantify the

relative fraction of impurities in the measured CNT sample

using the area ratio of D/G bands under fixed laser power den-

sity. In addition, the diameters and electronic structures of

CNTs can be determined by using the resonance Raman scat-

tering[30]. However, the drawback of Raman spectroscopy is

that it cannot provide direct information on the nature of

metal impurities, and it is not as effective in studying CNT

samples with a low content of amorphous carbon[31].

UVvis-NIR spectroscopy is a rapid and convenient tech-

nique to estimate the relative purity of bulk SWCNTs based

on the integrated intensity of S22 transitions compared with

that of a reference SWCNT sample [28]. It is convenient to

determine the concentration of SWCNTs dispersed in solu-

tion once the extinction coefficient of SWCNTs is known

[32]. On the other hand, SWCNTs give rise to a series of pre-

dictable electronic band transitions between van Hove singu-

larities in the density states of nanotubes (S11, S22, and M11),

therefore this technique is also used to analyze SWCNT types,

i.e., metallic or semiconducting [31,33,34], according to their

electronic structure. For small diameter SWCNTs individually

dispersed in solution with the assistance of surfactants or

DNA molecules, the (n, m) index assignment is also possible

from UVvis-NIR spectroscopy [33,34]. The drawback of this

method is the difficulty in repeatedly preparing the standard-

ized SWCNT film or solution and controlling film thickness or

solution concentration, making it difficult for quantification

analysis. Furthermore, it is not yet possible to provide an

absolute value of the purity of SWCNTs because there is no

100% pure standard SWCNT sample or accurate extinction

coefficient for SWCNTs.

Besides the above most commonly used techniques, X-ray

photoelectron spectroscopy (XPS) is often used to character-

ize functional groups on the walls of CNTs, and energy disper-

sive spectroscopy (EDS) is also used to semi-quantitatively

identify the metal content in CNT samples, especially for

Fig. 1 TEM images of (a) amorphous carbon and fullerene molecules on the surface of CNTs [26]; (b) metal nanoparticles

covered by amorphous carbon layer, (c) metal nanoparticles covered by graphitic carbon multi-layer (reproduced with

permission from[27], Copyright 2004 Amercian Chemical Society).

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5 2005


4/23

trace amounts. The major purity and quality assessment

techniques and their efficiency are summarized inTable 1.

It seems that no assessment technique mentioned above

can give a precise and comprehensive quantification of CNTs

(Table 1). Consequently, there is a need to develop an inte-

grated method by which the type, amount, and morphology

of CNT-containing materials can be accurately and precisely

quantified [35]. Alternatively, a combination of different

assessment techniques may be a good choice to give a full

understanding of CNTsbut this takes more time. Furthermore,

a precise definition of purity should be established because

purity can be different from different points of view, such

as CNT content, structure integrity, and SWCNT content. From

this respect, we define the purity of CNTs as given inTable 2.

Meanwhile, the major purity assessment techniques and

how to evaluate them are also briefly included.

1.4. Purpose of this review

As mentioned above, a series of problems involving the pres-

ence of impurities in CNTs, the non-uniformity in morphol-

ogy and structure of both CNTs and impurities, as well as

the absence of precise characterization methods limit the

applications of CNTs. Thus great attention has been paid to

the issue of purification. The developed purification schemes

usually take advantage of differences in the aspect ratio and

oxidation rate between CNTs and carbonaceous impurities.

In most cases, CNT purifications involve one or more of the

following steps: gas phase oxidation, wet chemical oxida-

tion/treatment, centrifugation, filtration, and chromatogra-

phy, etc. However, a reproducible and reliable purification

protocol with high selectivity, especially for SWCNTs, is still

a great challenge, because the purity of CNTs depends on

not only purification itself, but also many other factors,

including CNT type (SWCNTs or MWCNTs), morphology and

structure (defects, whether or not they exist in bundles, diam-

eter), impurity type and their morphology (particle size, de-

fect, curvature, the number and crystallinity of carbon

layers wrapping metal particles), purity assessment tech-

nique, and so on.

This article attempts to give a comprehensive survey and

analysis of the purification of CNTs. The challenges existing

in the purification methods, synthesis techniques and purity

assessments, which have to be overcome in order to enable

the wide applications of CNTs, will be discussed. The purity

in this article generally is referred to as CNT content in the

Table 1 Summary of commonly used techniques for detecting the impurities in CNT samples

Technique C-Ia M-Ib F-Gc S-Dd C-Fe Advantages Limitations

EM Df D g Jh Direct observation A small amount of sample is analyzed

TGA J J Precise content of carbon and metals CNTs analyzed are completely destroyed

Raman D D J Diameter, quality and conductivity of SWCNTs Invalid for MWCNTs and metal impurities

UVvis-NIR J J Conductivity feature and content of SWCNTs Need 100% pure SWCNTs as standard

XPS J Accurate assessment of F-G on CNTs Invalid to purity assessment

EDS J Elemental contents, special for trace amounts Invalid to evaluate CNT content

a Carbonaceous impurity.

b Metal impurity.

c Functional groups.

d Structure defects.

e Conductivity feature.

f Qualitatively valid.

g Invalid.

h Valid.

Table 2 Definition of purity for CNTs from different points of view and the corresponding assessment techniques

Purity definition Assessment technique and methods

CNT content The content of CNTs in sample containing CNTs,

carbonaceous and metallic impurities

TGA

Metal content can be calculated from the ash weight after

complete oxidation, and carbonaceous impurity content can be

calculated by corresponding peak area ratio from DTG curve.

CNTs without any other carbonaceous impurity are

characterized by one DTG peak

Structure integrity Pure CNTs without large defects and faults, and

no functional groups, amorphous carbon or

fullerene adhered on the tube wall

EM in combination with XPS

EM can directly observe and qualitatively assess the amount of

defects, amorphous carbons, fullerenes adhered on the wall of

CNTs. XPS can give a quantitative characterization of type and

content of functional groups

SWCNT conten t The content of SWCNTs in CNTs Raman spectroscopy

100% pure SWCNTs should be characterized by one G band

with RBM and without D band

2006 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


5/23

as-prepared or purified samples, and the yield means the

weight ratio of purified CNTs to that of the as-prepared CNT

sample, unless specified otherwise.

2. Purification methods

Purification methods of CNTs can be basically classified into

three categories, namely chemical, physical, and a combina-

tion of both. The chemical method purifies CNTs based on

the idea of selective oxidation, wherein carbonaceous impuri-

ties are oxidized at a faster rate than CNTs, and the dissolu-

tion of metallic impurities by acids. This method can

effectively remove amorphous carbon and metal particles ex-

cept for those encaged in polyhedral graphitic particles. How-

ever, the chemical method always influences the structure of

CNTs due to the oxidation involved. The physical method

separates CNTs from impurities based on the differences in

their physical size, aspect ratio, gravity, and magnetic proper-

ties, etc. In general, the physical method is used to remove

graphitic sheets, carbon nanospheres (CNSs), aggregates or

separate CNTs with different diameter/length ratios. In prin-

ciple, this method does not require oxidation, and therefore

prevents CNTs from severe damage. However, the physical

method is always complicated, time-consuming and less

effective. The third kind of purification combines the merits

of physical and chemical purification, and we denominate it

as multi-step purification in this article. This method can lead

to high yield and high-quality CNT products. Owing to the

diversity of the as-prepared CNT samples, such as CNT type,

CNT morphology and structure, as well as impurity type and

morphology, it needs a skillful combination of different puri-

fication techniques to obtain CNTs with desired purity.

2.1. Chemical oxidation

The carbonaceous impurities co-existing with as-synthesized

CNTs are mainly amorphous carbon and CNPs. Compared

with CNTs, these impurities usually have higher oxidation

activity. The high oxidative activity demonstrated by amor-

phous carbon is due to the presence of more dangling bonds

and structural defects which tend to be easily oxidized;

meanwhile the high reactivity of the CNPs can be attributed

to their large curvature and pentagonal carbon rings [36,37].

Therefore, chemical oxidation purification is based on the

idea of selective oxidation etching, wherein carbonaceous

impurities are oxidized at a faster rate than CNTs. In general,

chemical oxidation includes gas phase oxidation (using air,

O2, Cl2, H2O, etc.), liquid phase oxidation (acid treatment

and refluxing, etc.), and electrochemical oxidation. The disad-

vantages of this method are that it often opens the end of

CNTs, cuts CNTs, damages surface structure and introduces

oxygenated functional groups (OH, C@O, and COOH) on

CNTs. As a result, the purified CNTs in turn can serve as

chemical reactors or a starting point for subsequent nanotube

surface chemistry[38,39].

2.1.1. Gas phase oxidation

In gas phase oxidative purification, CNTs are purified by oxi-

dizing carbonaceous impurities at a temperature ranging

from 225 C to 760 C under an oxidizing atmosphere. The

commonly used oxidants for gas phase oxidation include air

[4046], a mixture of Cl2, H2O, and HCl[47], a mixture of Ar,

O2, and H2O [4850], a mixture of O2, SF6 and C2H2F4 [51],

H2S and O2[52], and steam[53].

High temperature oxidation in air is found to be an extre-

mely simple and successful strategy for purifying arc dis-

charge derived MWCNTs, which are metal free and have

fewer defects on tube walls. Ebbesen et al. [40,41] first re-

ported a gas phase purification to open and purify MWCNTs

by oxidizing the as-prepared sample in air at 750C for

30 min. However, only a limited amount of pure MWCNTs

(12 wt%) remained after the above purification, which can

be ascribed mainly to two reasons. One is uneven exposure

of CNTs to air during oxidation, and the other is the limited

oxidation selectivity between CNTs and carbonaceous impu-

rities. Therefore, two routes may be helpful to increase the

purification yield using this simple air oxidation. One is to en-

sure that the as-synthesized CNT samples are evenly exposed

to air, and the other is to enhance the difference in oxidation

resistance to air between CNTs and carbonaceous particles.

The above suggestions have been verified by some research-

ers. As an example, Park and coworkers [42] increased the

purification yield to 35 wt% by rotating the quartz tube in

which the sample was placed, in order to evenly expose the

CNTs and carbonaceous impurities to air at 760 C for 40 min.

To increase the difference in oxidation resistance to air be-

tween MWCNTs and carbon impurities, the difference in oxi-

dation rates of graphite and intercalated graphite[4345]was

taken into account. Graphite intercalation compounds are

formed by the insertion of atomic or molecular layers of other

chemical species between graphite layers. This interaction

causes an expansion of carbon interlayer spacing, which re-

duces the oxidation resistance of the intercalated graphite.

Carbonaceous impurities have higher structural defect densi-

ties than CNTs, and are therefore more ready to act as reac-

tion sites for intercalated atoms. Thus the oxidation

resistance difference between CNTs and carbonaceous impu-

rities can be increased. As an example, Chen et al. [43] re-

ported a combined purification process consisting of

bromination and subsequent selective oxidation with oxygen

at 530 C for 3 days. Temperature programmed oxidation pro-

files of the CNT samples with and without bromine treatment

are shown inFig. 2. It is obvious that oxidation of the bromi-

nated sample occurs more readily than that without bromin-

ation. TEM studies showed that CNTs with both ends open

were enriched in the purified sample, and the yield obtained

by the above process varied from 10 to 20 wt% with respect

to the weight of the original carbon sample. Furthermore,

they found that the yield depended crucially on the flow rate

of oxidant, the amount of initial sample, the manner of pack-

ing of the carbon, and the quality of the cathodic soot.

Although MWCNTs can be purified by a variety of gas

phase oxidation [4145], attempts to use similar procedures

for SWCNTs result in nanotubes etching away. For example,

using the bromine and oxygen system, the yield was

3 wt% [47] for SWCNT purification, which implies that a

large fraction of SWCNTs are consumed in the process. This

large difference between MWCNTs and SWCNTs results from

two factors. One is the larger amount of curvature experi-

enced by the graphene sheet of SWCNTs, and the other is

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5 2007


6/23

metal impurities catalyzing the low-temperature oxidation of

carbon. There may therefore be two ways to increase the puri-

fication yield of SWCNTs using gas phase oxidation. One is to

select oxidants that can selectively oxidize carbonaceous

impurities by a unique selective carbon surface chemistry

while leaving SWCNTs intact. The other is to remove metal

particles before gas phase oxidation. Some positive results

have been obtained following the above suggestions.

Zimmerman et al. [47] first reported suitable conditions

allowing for the removal of amorphous or spherical carbon

particles, with or without metal catalyst inside, while simul-

taneously protecting SWCNTs. The purification incorporates

a chlorine, water, and hydrogen chloride gas mixture to re-

move the impurities. A SWCNTyield of 15 wt% and a purity

of90% indicate that the carbonaceous impurities are prefer-

entially removed. Based on their experimental observation,

hydrogen chloride was required for selective removal of the

unwanted carbon. They proposed a mechanism for the puri-

fication. Chlorine gas mixture interacted with the nanotube

cap and formed a hydroxy-chloride-functionalized nanotube

cap. Hydrogen chloride in the gas phase purification mixture

protected the caps that are more reactive, by preventing hy-

droxyl groups from deprotonating. The disadvantage of this

method is that only small quantities (5 mg) of SWCNTs were

purified each time. Furthermore, the reagents and produced

gases are toxic and explosive, which limits its practical use.

At the same time, some other oxidants that can selectively

oxidize carbonaceous impurities were also reported. For

example, hydrogen sulfide was reported to play a role in

enhancing the removal of carbon particles as well as control-

ling the oxidation rate of carbon. A purity of 95% SWCNTs

with a yield of 2050 wt% depending on the purity of raw

material was reported[52]. In addition, steam at 1 atm pres-

sure [53], local microwave heating in air [46], air oxidation

and acid washing followed by hydrogen treatment[54]were

also reported to work well to improve the purification yield.

It was Chiang et al.[48,49]who clearly elucidate the role of

metals in oxidizing carbons and the need for their prior re-

moval. They found that metal particles catalyze the oxidation

of carbons indiscriminately, destroying SWCNTs in the pres-

ence of oxygen and other oxidizing gases. Encapsulated metal

particles can be exposed using wet Ar/O2(or wet air) oxidation

at 225 C for 18 h. This exposure was attributed to the expan-

sion of the particles because oxidation products have a much

lower density (the densities of Fe and Fe2O3 are 7.86 and

5.18 g/cm3, respectively). Such significant expansion broke

the carbon shells, and the particles were exposed as a result.

Based on the above results, they proposed a multi-stage proce-

dure for purifying SWCNTs synthesized by the HiPco process.

Their method begins with cracking of the carbonaceous shells

encapsulating metal particles using wet oxygen (20% O2in ar-

gon passed through a water-filled bubbler) at 225 C, followed

by stirring in concentrated hydrochloric acid (HCl) to dissolve

the iron particles. After filtering and drying, the oxidation

and acid extraction cycle was repeated once more at 325 C,

followed by an oxidative baking at 425 C. Finally, 99.9% pure

SWCNTs (with respect to metal content) with a yield of

30 wt% were obtained. The validity of this method was veri-

fied by another group[50]. However, owing to the complicated

purification steps, it is hard to purify SWCNTs in a large scale.

Xu et al. [51] developed a controlled and scalable multi-

step method to remove metal catalyst and non-nanotube car-

bons from raw HiPco SWCNTs. Their scalable multi-step puri-

fication included two processes: oxidation and deactivation of

metal oxides. In the oxidation, metal catalysts coated by non-

nanotube carbon were oxidized into oxides by O2and exposed

by using a multi-step temperature increase program. In the

deactivation step, the exposed metal oxides were deactivated

by conversion to metal fluorides through reacting with

C2H2F4, SF6, or other fluorine-containing gases to avoid the

catalytic effect of iron oxide on SWCNToxidation. The Fe con-

tent was remarkably decreased from 30 to 1 wt% and a

yield of2548 wt% was achieved. However, the shortcoming

of this method is that it is limited to HiPco SWCNTs, in which

the dominant impurity is metal catalyst. Furthermore, the

toxicity of the reagents used in this method and the resulting

gases are undesirable features.

Gas phase oxidation is a simple method for removing car-

bonaceous impurities and opening the caps of CNTs without

vigorously introducing sidewall defects, although it cannot di-

rectly get rid of metal catalyst and large graphite particles.

Thus it is a good choice to purify arc discharge derived

MWCNTs, which contains no metal catalyst. For purifying

SWCNTs or MWCNTs (synthesized by other techniques), acid

treatment to remove the metal catalyst is always necessary.

Another point worth noting is that CNTs (SWCNTs in particu-

lar) in agglomerates prevent oxidant gas from homogeneously

contacting the whole sample. In order to obtain high-purity

CNTs, the amount of sample to be purified each time is quite

limited (tens to a hundred milligrams). Therefore, methods

that can cause the oxidant gas to homogeneously contact

CNT samples are urgently required to obtain high-purity CNTs

on a large scale. Recently, Tan et al.[55]mixed raw SWCNTs

with zirconia beads to enhance air flow uniformity and in-

crease the exposed surface of raw soot during thermal oxida-

tion in air. The final purified samples had a yield of 26 wt%

and a metal impurity of7%. Although the purity is not very

high, the technique suggests a way to purify SWCNTs on a

large scale using gasphase oxidation. This method canprovide

pure and opened CNTs without heavily damaging tube walls,

which is a good choice for theapplication of open-ended CNTs

as nano-size reaction tubes or chemical reactors[56,57]. For

achieving purified CNTs on a large scale, gas phase oxidation

Fig. 2 Temperature programmed oxidation profiles of the

cathodic soot before (CS) and after (BS) bromination

(reprinted with permission from[43], Copyright 1996 Wiley

VCH Verlag GmbH & Co. KGaA).

2008 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


7/23

need to be modified in the following ways: one is to look for a

simple approach and non-toxic reagents to remove metal par-

ticles encapsulated by carbon layers; the other is to look for a

way that can make oxidant gas homogeneously contact the

as-prepared CNTs. In addition, the gas phase oxidation can

be combined with other techniques, such as filtration or cen-

trifugation, to further enhance the purification efficiency.

2.1.2. Liquid phase oxidation

Although the merits of gas phase oxidation are obvious, it has

a drawback that metal particles cannot be directly removed,

and further acid treatment is needed. In order to overcome

this limitation, liquid phase purification that always simulta-

neously removes both amorphous carbon and metal catalyst

was developed. Oxidative ions and acid ions dissolved in solu-

tion can evenly attack the network of raw samples, and there-

fore selection of oxidant type and precise control of treatment

condition can produce high-purity CNTs in a high yield. The

commonly used oxidants for liquid phase oxidation include

HNO3 [5860], H2O2 or a mixture of H2O2 and HCl [6163], a

mixture of H2SO4, HNO3, KMnO4 and NaOH [6467], and

KMnO4 [6769]. The shortcomings of this method are that it

causes reaction products on the surface of CNTs, adds func-

tional groups, and destroys CNT structures (including cutting

and opening CNTs).

Nitric acid is the most commonly used reagent for SWCNT

purification for its mild oxidation ability, which can selec-

tively remove amorphous carbon. In addition, it is inexpen-

sive and nontoxic, capable of removing metal catalysts and

no secondary impurities are introduced.

Dujardin et al.[58]reported a one-step method using con-

centric nitric acid to purify SWCNTs synthesized by laser abla-

tion. Briefly, as-synthesized SWCNTs were sonicated in

concentrated nitric acid for a few minutes followed by reflux-

ing under magnetic stirring at 120130 C for 4 h. The yield

reached 3050 wt% of the raw sample and the metal amount

was decreased to 1 wt%. One problem in the above purifica-

tion is that the permeation rate during filtration was very

low because SWCNTs packed together and the filter mem-

brane was blocked. This makes it difficult to purify CNTs on

a large scale, and some small carbonaceous impurity particles

cannot permeate the filter. To solve this problem, Rinzler et al.

[59]adopted hollow-fiber cross-flow filtration (CFF) to filtrate

SWCNTs that had been refluxed in 2.6 M HNO3for 45 h. Highly

pure SWCNTs with a yield of 1020 wt% were obtained with

this readily scalable method, which opens up a way to purify

SWCNTs on a large scale. Even though the effectiveness of ni-

tric acid treatment on the purification of SWCNTs is con-

firmed, the relationship between purification yield and purity

with systematic and quantitative measurements was not re-

ported before Hu et al.s work[60]. They established a system-

atic and quantitative relationship between yield and purity by

using solution phase NIR spectroscopy. In their experiments,

1 g of the as-prepared SWCNT sample was refluxed in 3 M ni-

tric acid for 12, 24 and 48 h, in 7 M nitric acid for 6 and 12 h,

and in concentrated nitric acid for 6 and 12 h. The weight per-

cent of each component calculated from TGA and NIR spectra

is plotted inFig. 3. It is clear that the purity and the yield of

SWCNTs with nitric acid treatment depend on the concentra-

tion of the nitric acid and the time of reflux. The treatments of

3 M HNO3 for 12 hand7 M HNO3 for6 hwere the most efficient.

Nitric acid treatment destroys SWCNTs, leading to the produc-

tion of carbonaceous impurities. Nevertheless, with the ability

to dissolve the metal catalyst, intercalate SWCNT bundles, at-

tack amorphous carbon, and break large carbon particles, the

nitric acid treatment can be a viable first step for SWCNT puri-

fication. The key to achieving high-purity SWCNTs is a subse-

quent process for removing the carbonaceous impurities that

remain in the sample after nitric acid treatment. In this case,

a preferred step is hollow-fiber CFF[59].

Hydrogen peroxide (H2O2) is also a mild, inexpensive and

green oxidant, which can attackthe carbonsurface.The disad-

vantage of H2O2is also obvious. It cannot remove metal parti-

cles. Therefore, it is usually used together with HCl. HCl is a

widespread chemical that can be easilyconverted into a harm-

less salt. Therefore,purifying CNTs using H2O2 followedby HCl

treatment to remove metal particles has also been intensely

investigated. Macro-scale purification, including a first reflux-

ing treatment in H2O2 solution and then rinsing with HCl,

was reported by Zhao et al.[61,62]. Their experimental results

showed that the size of Fe particles has a great influence on

the oxidation of amorphous carbon. However, this was still a

question about the effect of Fe before Wangs work[63].

Wang et al.[63]tried to explain the above question. They

combined two known reactions (oxidation of amorphous car-

bon with H2O2and removal of metal particles with HCl) into a

single pot, which simplified the process. Surprisingly, the

product yield and purity were improved. Typically, carbon-

coated iron impurities were simply dissolved by reacting with

an aqueous mixture of H2O2 and HCl at 4070 C for 48 h.

With this treatment, the purification yield was significantly

increased to 50 wt% and the purity was up to 96 wt%.

According to Wang, the effect of this process on the purifica-

tion can be summarized as following. First, Fe particles act as

a catalyst by Fenton chemistry[70], producing hydroxyl radi-

cals (OH), a more powerful oxidant than H2O2. Second, HCl

dissolves the iron nanoparticles upon their exposure. The ex-

posed iron releases ferrous ions as a result of dissolution of

Fig. 3 Mass balance of the normalized weight percentage of

all components including SWCNTs, metal, carbonaceous

impurities, and weight loss of the SWCNT samples

(reprinted with permission from[60], Copyright 2003

Amercian Chemical Society).

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5 2009


8/23

the Fe particles in the acid solution. The ferrous ions quickly

diffuse into the acid solution, thereby eliminating iron and

iron hydroxide precipitation and their unwanted catalytic ef-

fect (Fig. 4). Therefore, by confining the catalytic effect to the

vicinity of the carbon-coated iron nanoparticles, both a high

selectivity in removing iron particles and carbonaceous shells

and a low consumption of SWCNTs are accomplished.

At almost the same time, microwave-assisted inorganic

acid treatment for the effective removal of metal particles

was reported [7176]. The principle of this method is that

inorganic acids such as HNO3, HCl, and H2SO4can rapidly ab-

sorb microwave energy and dissolve metals efficiently with-

out damaging the tube wall structure in a short time.

As discussed above, HNO3, H2O2, as well as microwave-as-

sisted inorganic acid treatments can effectively remove metal

particles, but they are not so effective in removing carbona-

ceous particles owing to the relative mildness in their oxida-

tion. In order to get rid of carbonaceous impurities, liquid

oxidants with stronger oxidation activity were also investi-

gated. These oxidants are predominantly mixture of acids

and KMnO4.

Liu et al. [64]use a mixture of concentrated H2SO4/HNO3(3:1 by volume) to cut highly tangled long ropes of SWCNTs

into short, open-ended pipes, and thus produced many car-

boxylic acid groups at the open ends. Wiltshire et al. [65]re-

ported that liquid phase oxidation could be a continuous

diameter-selective process, eliminating SWCNTs with smaller

diameter by oxidizing the sidewalls. Li and coworkers [66]

investigated the purification effectiveness of concentrated

H2SO4/HNO3 (3:1) treatment and compared this with 6 M

HNO3 treatment. Typical TEM images of purified SWCNTs

after different treatment conditions are shown inFig. 5, from

which it can be concluded that concentrated H2SO4/HNO3(3:1) is more effective than nitric acid in removing impurities.

Furthermore, it was reported that the best purification condi-

tion could reach 98% purity of SWCNTs with a yield of 40 wt%

within 2 h, without decreasing the number of small diameter

nanotubes for a 3 h reflux process using a concentrated

H2SO4/HNO3mixture (3:1).

Colomer et al.[68]reported an effective method for remov-

ing amorphous carbon by refluxing as-prepared MWCNTs in

acidified KMnO4 at low-temperature (80C). According to

them, amorphous carbon was completely removed at the cost

of more than 60% carbon loss. TEM observation of the purified

CNTs indicated that all amorphous carbon aggregates were

removed and the CNT caps were opened. Hernadi et al. [69]

verified the above conclusion. They obtained MWCNTs with

oxygen functional groups which were free from amorphous

carbon by KMnO4 oxidation. Zhang et al. [67] investigated

the effect of KMnO4 in alkali solution on the purification of

SWCNTs. KMnO4in alkali solution is a much more moderate

Fig. 5 TEM images of purified SWCNTs: (a) sonication in 6 M HNO 3for 4 h, (b) refluxing in 6 M HNO3at 120 C for 4 h,

(c) refluxing in concentrated H2SO4/HNO3mixture (3:1) at 120 C for 2 h, (d) refluxing in concentrated H2SO4/HNO3mixture

at 120 C for 4 h (reprinted with permission from [66], Copyright 2004 Institute of Physics Publishing).

Fig. 4 Scheme of localized catalytic reaction of H2O2with

carbon-coated iron nanoparticles (not drawn to scale,

reprinted with permission from[63], Copyright 2007

Amercian Chemical Society).

2010 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


9/23

oxidant than in acidic solution. The solution cannot effec-

tively open the tube, while it is strong enough to attack the

nanotube walls and generate abundant functional groups.

The problem of this process is that additional steps are

needed to remove the MnO2generated during the oxidation.

It is desirable to remove carbonaceous impurities by convert-

ing them into soluble or volatile products, and from this point

of view, KMnO4seems to be a less suitable oxidation agent.

Liquid phase oxidation is a continuous process that can

eliminate impurities on a large scale, and it is hoped that it

can be widely used for industrial application in the future.

This method often leads to surface modification that prefer-

entially takes place on CNT sidewalls, which increases the

chemical activity and the solubility of CNTs in most organic

and aqueous solvents. This surface modification effect shows

great potential for improving their physical and chemical

properties for specific applications, e.g., in making mechani-

cally reinforced composites, in use as scanning probe micros-

copy tips with tailored chemical sensitivity, and in producing

nanotube derivatives with altered electronic structures and

properties [7780]. Furthermore, CNTs can be cut into short

fragments decorated with oxygen functional groups under

suitable treatment conditions, which greatly increases their

dispersibility and facilitates their practical applications. For

example, the application of CNTs in the field of emerging bio-

technology is based on the premise that short CNTs are dis-

persible in water [81,82]. The main problem of this liquid

oxidation strategy is the damage to CNTs, the inability to re-

move large graphite particles, and the loss of a large amount

of SWCNTs with small diameter. It is very difficult to obtain

purified SWCNTs with high-purity and high yield without

damage by simply using liquid phase oxidation.

2.1.3. Electrochemical oxidation

As with liquid phase oxidation and gas phase oxidation, car-

bon materials with fewer defects usually show a lower corro-

sion rate under electrochemical oxidation. Therefore, it is

reasonable to deduce that CNTs with fewer defects should

show higher electrochemical oxidation resistance than car-

bon impurities with more defects.

Fang et al.[27]investigated the electrochemical cyclic vol-

tammetric (CV) oxidation behavior of an arc discharge derived

SWCNT sample in KOH solution. Amorphous carbon in the

as-grown SWCNT sample was effectively removed by the CV

oxidation, as confirmed by analyzing the sp3/sp2 carbon ratio

from C1s XPS spectra and TEM observations. The removal of

amorphous carbon led to the exposure of metal nanoparti-

cles, hence facilitating the elimination of the metal impurities

by subsequent HCl washing. The redox peaks from the elec-

trochemical redox reactions of Fe and Ni impurities can be

considered as an indication of the extent of removal of the

amorphous carbon, and the optimum electrochemical oxida-

tion time for the purification of the as-grown SWCNT sample

can be determined in real time during the CV oxidation

treatment.

The above electrochemical oxidation was performed in

KOH solution, which needs further acid treatment to remove

metal particles. This makes the purification complex. If the

solution is acidic, the post-treatment should be omitted,

which makes the purification easier. Ye et al.[83]verified this.

They recently reported an ultra-fast and complete opening

and purification of MWCNTs through electrochemical oxida-

tion in acid solution. The vertically aligned MWCNT (with her-

ringbone structure) arrays investigated were grown on a

carbon microfiber network through DC plasma-enhanced

CVD. Electrochemical oxidation for tip opening and purifica-

tion of MWCNT arrays was performed in an aqueous solution

of 57% H2SO4at room temperature. SEM and TEM images be-

fore and after purification (Fig. 6) indicated that the CNT tips

were opened, and entrapped metals were removed during the

electrochemical oxidation. The results of inductively coupled

plasma-mass spectrometry indicate that 98.8% of the Ni was

removed after the electrochemical oxidation in acid. The

authors also investigated a series of electrolyte solutions for

electrochemical opening of CNT tips at room temperature.

They concluded that if electrochemical oxidation was per-

formed in neutral or basic aqueous solutions, no significant

tip opening was observed. If aqueous solutions of common

strong or medium strength acids (5% H2SO4, 5% HNO3, or

25% HNO3+ 25% H2SO4, 5% H3PO4 and 5% CH3COOH) were

used, not only can the amorphous carbon be readily etched

but also the metal catalyst can be dissolved.

Superior to the gas phase oxidation and wet oxidation, the

optimum time and degree of electrochemical oxidation for

CNT purification can be easily determined. This method can

get rid of impurities to some extent, particularly for selec-

tively opening and purifying vertically aligned CNT arrays.

The desired vertical orientation can be maintained and facil-

itates the use of CNT arrays as fuel cell electrodes, sensor

platforms, nanoreactors, field emitter components, and other

applications. However, little polyhedral carbon, graphite par-

ticles, and metal particles enwrapped by carbon layers with

fewer defects can be removed by the CV oxidation. Moreover,

the purity of the obtained sample greatly depends on the

starting materials, and the amount of sample purified for

each batch is too small to make the method practical.

2.1.4. Brief summary

Chemical-based purification can effectively remove amor-

phous carbon, polyhedral carbon, and metal impurities at

the expense of losing a considerable amount of CNTs or

destroying CNT structures. Gas phase purification is charac-

terized by opening the caps of CNTs without greatly increas-

ing sidewall defects or functional groups. Liquid phase

oxidation introduces functional groups and defects preferen-

tially on CNT side walls, and may cut CNTs into shorter ones

with different lengths. The electrochemical oxidation is suit-

able for purifying CNT arrays without destroying their align-

ment. These features allow chemical purification adopted by

researchers to fulfill different requirements. The most serious

problem of this technique is that the structure of CNTs may

be destroyed by the reactants, and hence limits the applica-

tions of CNTs in some fields, for example, electronic devices.

2.2. Physical-based purification

A big problem in chemical purification is that it always de-

stroys the structure of CNTs or changes their natural surface

properties. To elucidate the inherent physical and chemical

properties of CNTs, purifications that do not involve oxidative

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5 2011


10/23

treatment are highly desirable. The morphology and physical

properties of CNTs, such as aspect ratio, physical size, solubil-

ity, gravity, and magnetism are different from impurities.

These differences enable one to separate CNTs from impuri-

ties by adopting some physical techniques. Therefore, physi-

cal-based methods including filtration, chromatography,

centrifugation, electrophoresis, and high temperature (1400

2800 C) annealing, have been extensively investigated. The

most striking feature of these methods is a non-destructive

and non-oxidizing treatment. Another feature is that the

purifications are mostly performed in solution, which re-

quires the as-prepared samples to have a good dispersibility

in the solutions. To meet this requirement, surfactants and/

or sonication are often used.

2.2.1. Filtration

Separation by filtration is based on the differences in physical

size, aspect ratio, and solubility of SWCNTs, CNSs, metal par-

ticles and polyaromatic carbons or fullerenes. Small size par-

ticles or soluble objects in solution can be filtered out, and

SWCNTs with large aspect ratio will remain. Polyaromatic

carbons or fullerenes are soluble in some organic solvents,

such as CS2, toluene, etc. The impurities can be easily re-

moved by immersing the as-prepared sample in these organic

solutions followed by filtering. The impurity particles smaller

than that of the filter holes flow out with the solution during

filtration, while large impurity particles and small ones

adhering to the CNT walls remain. One problem of this tech-

nique is that CNTs or large particles deposited on the filter of-

ten block the filter, making the filtering prohibitively slow and

inefficient. Therefore, a stable suspension of CNTs and a tech-

nique preventing them from deposition and aggregation are

very important during filtration. Thus, surfactants are widely

used to make a stable CNT suspension, and ultrasonication is

usually adopted to prevent the filter from being blocked.

Bonard et al. [84] first applied filtration assisted with sonica-

tion to purify MWCNTs. The as-prepared MWCNTs were dis-

persed in water with sodium dodecyl sulfate (SDS), and a

stabilized colloidal suspension was formed. The suspension

was filtered using a filtration apparatus with a funnel large en-

ough to allow the sonication of the colloidal suspension to

extract larger particles. In order to enhance the separa-

tion yield, successive filtrations were carried out until the de-

sired purity is reached. Shelimov et al. [85] used the above

procedure to purify SWCNTs and obtained SWCNT material

with a purity of more than 90% (estimated by EM) and a yield

of 3070 wt%.

Bandow et al.[86]developed a purification process (shown

inFig. 7) consisting of filtration and microfiltration under an

overpressure (2 atm) of N2to separate CNSs, metal nanopar-

ticles, polyaromatic carbons and fullerenes from SWCNTs.

The microfiltration was repeated three times to minimize

the amount of residual CNSs and metal particles. Using this

technique, 84, 10, and 6 wt% of purified SWCNTs, CNSs,

and CS2extracts were separated from the as-prepared laser-

synthesized SWCNTs.

A major advantage of filtration is that it is driven by pure

physicochemical interactions of carbon products with amphi-

philic molecules and the filter membrane, leaving the nano-

tubes undamaged. However, this procedure relies on the

quality of raw samples and is time-consuming. In addition,

Fig. 6 SEM (a, b) and TEM (c, d) images of MWCNT arrays: (a, c) as-grown, (b, d) purified (reprinted with permission from

[83], Copyright 2006 Amercian Chemical Society).

2012 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


11/23

amorphous and spherical carbon particles stuck on the tube

walls cannot be effectively removed.

2.2.2. Centrifugation

Centrifugation is based on the effect of gravity on particles

(including macromolecules) in suspension because two parti-

cles of different masses settle in a tube at different rates in re-

sponse to gravity. On the other hand, centrifugation can also

separate amorphous carbon and CNPs based on the different

stabilities in dispersions consisting of amorphous carbon,

SWCNTs, and CNPs in aqueous media. The different stabili-

ties resulted from the different (negative) surface charges

introduced by acid treatment [87,88]. Low-speed centrifuga-

tion (2000g) is effective in removing amorphous carbon and

leaving SWCNTs and CNPs in the sediment. High-speed cen-

trifugation (20000g) works well in settling CNPs, while leaving

SWCNTs suspended in aqueous media. The effectiveness of

centrifugation in separating SWCNTs from amorphous car-

bon and CNPs is shown inFig. 8. The drawback of this process

is that CNTs need to be first treated with nitric acid, which

introduces functional groups on their surface.

2.2.3. Solubilization of CNTs with functional groups

The principle of this purification step is to solubilize CNTs by

introducing functional groups onto their surface. These solu-

ble nanotubes allow for the application of other techniques

such as filtration or chromatography as a means of tube puri-

fication. To regain reasonable quantities of un-functionalized

but purified nanotubes, the functional groups should be re-

moved by thermal treatment or other techniques.

Coleman et al. [89,90] described a one-step, high yield,

nondestructive purification for MWCNTs containing soot

using a conjugated organic polymer host (poly(m-pheny-

lene-co-2,5-dioctoxy-p-phenylenevinylene (PmPV)) in tolu-

ene. PmPV is shown to be capable of suspending nanotubes

indefinitely whilst the accompanying graphitic particles settle

out. Finally the host polymer was removed by Buchner filtra-

tion, giving CNTs with a purity of 91% (estimated from elec-

tron paramagnetic resonance). In this case, a yield of 17 wt%

pristine nanotubes was reclaimed from the soot.

Yudasaka et al. [91] mixed as-grown SWCNTs with a 2%

monochlorobenzene (MCB) solution of polymethylmethacry-

late (PMMA) with an ultrasonic cleaner. The mixture was

homogenized through an ultrasonic-homogenizer and fil-

tered. The MCB was removed by evaporation at 150 C, and

PMMA was removed by burning it off in 200-Torr oxygen gas

at 350 C. At the same time, azomethine ylides and solution

phase ozonolysis (78 C) were also reported to solubilize

CNTs via 1,3 dipolar cycloaddition[9294].

Recently, Jeynes et al.[95]and Sanchez-Pomales et al.[96]

reported a method for purifying CNTs using RNA and DNA.

Briefly, arc discharge derived CNTs were sonicated in deion-

ized water at 0 C (in an ice-water bath) for 30 min with

0.5 mg/mL total cellular RNA. The solution was then centri-

fuged to pellet the insoluble particles. RNA-wrapped CNTs

were treated with enzyme ribonuclease to remove the RNA

and thereby precipitate the CNTs. Jeynes et al. [95]also sug-

gested that RNA/DNA was more efficient in solubilizing CNTs

than SDS as there is a large surface area of phosphate back-

bone which interacts with water, while similarly there are

many bases to bind the CNTs.

The advantage of this process is that it can always pre-

serve the surface electronic structure of CNTs. This property

Fig. 8 TEM images of (a) SWCNT-COOH material showing embedded catalyst particles, (b) purified SWCNT-COOH fraction,

and (c) carbon particle fraction (reprinted with permission from[88], Copyright 2006 Amercian Chemical Society).

ExtractSolids caught

on filter

As-prepared carbonaceous sample in CS2

(SWCNT, CNS, C60, C70, polyaromatic carbons)

Evaporate CS2

Fullerenes (C60, C70),

polyaromatic carbons

Evaporate and

collect solids

Sonication in aqueous

solutions (0.1% surfactant)

CNSs

Filtration

Purified SWCNTs

Microfiltration under

over pressure (~ 2 atm)

Liquid

Solids caught

on filter

Fig. 7 A diagram illustrating the technique used forseparating coexisting CNSs, metal nanoparticles, and

polyaromatic carbons or fullerenes from the

laser-synthesized SWCNTs (modified with permission

from[86], Copyright 1997 Amercian Chemical Society).

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5 2013


12/23

is of fundamental importance for the use of nanotubes as bio-

sensors[97]. On the other hand, the capability of dispersing

CNTs in solution is a very important step for using CNTs as

vectors to deliver therapeutics (drug, nucleic acid) [97]. How-

ever, the effectiveness of this technique is not high for CNT

samples containing a large amount of impurities or bundled

SWCNTs.

2.2.4. High temperature annealing

For some applications of CNTs, such as their use as bio-mate-

rials, complete removal of metal particles is of particular

importance. However, it is very difficult to achieve this by acid

washing because most of the metal particles are enwrapped

by carbon layers. The physical properties of carbon and met-

als are different at high temperature (>1400 C) under inert

atmosphere or high vacuum. It is well known that, graphite

is stable even at 3000 C, while metal evaporates at tempera-

tures higher than their evaporation point. Therefore, it is ex-

pected that high temperature annealing can effectively

remove metal particles.

Lambert et al.[98]first attempted to remove metal catalyst

particles from SWCNTs by heating the material above the

evaporation temperature of the metal. The results showed

that this might be a good way to eliminate the catalyst parti-

cles. The effectiveness of removing metal particles in

MWCNTs using high temperature annealing was also verified

by a few reports[99101]. Their results suggest that high-pur-

ity (99.9%) CNTs with respect to metal particles can be ob-

tained by a high temperature (P1800 C) annealing

treatment. Further research indicated that high temperature

annealing (>1400 C) can change the structure of CNTs, such

as removing structural defects [102], enlarging diameter

[103], transforming SWCNTs to MWCNTs [104] or MWCNTs

to double-walled CNTs (DWCNTs) [105] at appropriate

temperatures.

In brief, high temperature annealing of CNTs is one of the

most efficient methods for the removal of metal particles at

the tips or in the hollow core of CNTs [98105] and also for

structural evolvement from disordered to straight, crystalline

layers [106]. High temperature annealing not only increases

the mechanical strength and thermal stability of CNTs but

also affects their electronic transport property. The drawback

of this method is that carbonaceous impurities still exist and

become more difficult to remove after graphitization. There-

fore, this method can be used to remove residual metal parti-

cles of purified CNTs obtained by other techniques for

achieving metal free CNTs. It can also be used to remove

the metal particles in as-prepared CNT samples that contain

a small amount of carbonaceous impurities or in the case

where the existence of carbon impurities is not of much

concern.

2.2.5. Other physical techniques

Some other physical methods were also explored to remove

metal particles, including a magnetophoretic technique

[107], supercritical fluid carbon dioxide (Sc-CO2) extraction

[108], and a mechanically ejecting technique [109]. These

techniques are reported to be effective in removing metal par-

ticles entrapped by carbon layers without changing the inher-

ent properties of the CNTs.

Kang and Park[107]demonstrated magnetophoretic purifi-

cation of SWCNTs (produced by HiPco process) from super-

paramagnetic iron-catalyst impurities in a microfluidic

device. The flow of a fluid through a microfluidic channel is

completely laminar and no turbulence occurs due to the

small dimension of the microchannels. By employing micro-

fluidic and a magnetic field-induced saw-tooth nickel micro-

structure, a highly enhanced magnetic force in adjoining

microchannels was exploited. The iron impurities of SWCNTs

were attracted towards areas of higher magnetic-flux density

in the microchannels where magnetic field was asymmetri-

cally generated perpendicular to the streamline. SWCNTs

with a purity of 9899% with respect to metal content were

obtained. However, some SWCNTs containing iron particles

were also removed, which decreased the yield of this method.

Sc-CO2has both gas-like and liquid-like properties; thus it

can penetrate into small pores like a gas and dissolve organic

substances like a liquid. Based on the above mechanism,

Wang and coworkers[108]developed a two-step purification

using Sc-CO2as a solvent to clean metal impurities from as-

grown SWCNTs produced by the HiPco process. The first step

of this method is a pretreatment procedure using bulk elec-

trolysis with ethylenediaminetetraacetic acid. The second

step is an in situ chelation/(Sc-CO2) extraction to remove me-

tal particles. Over 98% of the iron impurity (measured by EDS)

in the as-grown SWCNTs were removed using this two-step

extraction.

Thien-Nga et al.[109]developed a mechanical purification

to remove ferromagnetic particles from their graphitic shells.

The basic principle of the method is like a snooker game,

where the energy of elastic impact between encapsulated cat-

alysts and small hard inorganic particles is used to eject metal

kernels and trap them by a strong magnet. Typically, SWCNTs

were first dispersed in various solvents. Insoluble nanoparti-

cles (zirconium oxide, diamond, ammonium chloride, or cal-

cium carbonate) in the given medium were then added to

the suspension. This slurry was sonicated typically for 24 h.

This process enables the production of laboratory quantities

of SWCNTs containing no magnetic impurities.

2.2.6. Combination of purification and separation

Following the purification of CNTs to remove foreign materi-

als such as catalyst, amorphous carbon, and carbon-coated

nanoparticles, the sorting of SWCNTs according to their

length becomes particularly important in light of their poten-

tial applications. Various techniques have been employed to

purify CNTs and simultaneously sort them by length [110

113]. Chromatography is useful for the length fractionation

of shortened CNTs less than 300 nm in length [110,111]. For

longer CNTs, techniques such as capillary electrophoresis

(CE)[112]and field-flow fractionation (FFF) appear to be more

applicable [113]. These techniques are simple and non-

destructive for the purification and length-dependent separa-

tion of CNTs. These techniques require CNTs to be purified

are individually dispersed. Therefore, complex procedures

are needed to obtain this kind of highly dispersible CNT

solution.

Chromatography is a separation method that relies on dif-

ferences in partitioning behavior between a flowing mobile

phase and a stationary phase to separate the components

2014 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


13/23

in a mixture. Materials that are smaller than the pore size can

enter the pores and therefore have a longer path and longer

transit time, while larger materials that cannot enter the

pores are eluted first. Different molecules, therefore, have dif-

ferent total transit times through the column depending on

their size and shape. Chromatography was first used to sepa-

rate CNTs from carbonaceous impurities[114117]. After this,

chromatography is usually used to separate CNTs by length.

For example, purification and length separation of oxidatively

shortened SWCNTs were achieved by this technique in an

alkalescent water solution [110]and SWCNTs with different

lengths were separated. Huang and coworkers [111] used

chromatography to purify DNA-wrapped CNTs and sort them

into fractions of uniform length. As observed by atomic force

microscopy (AFM), the length variation was typically within

10% or less for each of the measured fractions (Fig. 9).

Electrophoresis is caused by electrostatic forces, which are

generated by applying an alternating current (AC) or DC elec-

tric field between an electrode and a charged body. Electro-

phoresis can be achieved regardless of the electric fields

uniformity. A charged particle is pulled along the field lines

toward the electrode carrying an opposite charge to that of

the particle. In the same field, a neutral body is merely polar-

ized. The result may produce a torque, but not a net transla-

tional force, without which the body as a whole will not move

towards either electrode. Electrophoresis can be appreciable

even when the free charge per unit weight of the particle is

quite small. Therefore, it is possible to purify CNTs in an elec-

tric field by using the motion difference between CNTs and

carbon impurities [118]. This motion depends not only on

the intrinsic electric properties but also on the diameter and

length of CNTs. Therefore, this technique can also be used

to separate CNTs with length, diameter, and conductivity by

refining experimental conditions. Yamamoto et al.[119]first

reported a purification and orientation method based on AC

electrophoresis in isopropyl alcohol. To increase separation

rate and effectiveness, Doorn et al.[112]adopted CE to purify

and separate CNTs by size. The CE was performed in narrow

tubes (in the order of lm) and resulted in rapid separation

based on charge- and size-dependent mobility of solution

phase species under the influence of an applied electric field.

And AFM observations on fractions demonstrated a length-

based separation mechanism that leads to elution of short

tubes first, followed progressively by longer tubes. Further

work [120] indicated that CE has the potential to separate

CNTs, not only by differences in length[112]but also by differ-

ences in size or other geometric factors, such as diameter or

cross section.

At the same time, AC dielectrophoresis was reported to be

capable of separating metallic SWCNTs from semiconducting

SWCNTs in SDS suspension[121]. This method takes advan-

tage of the difference in the relative dielectric constants of

two species with respect to the solvent, resulting in an oppo-

site movement of metallic and semiconducting tubes along

the electric field gradient. Metallic tubes are attracted toward

a microelectrode array, leaving semiconducting tubes in the

solvent. An enrichment of metallic tubes up to 80% was

achieved by a comparative Raman spectroscopy study on

the dielectrophoretically deposited tubes and a reference

sample.

In addition, FFF was also developed to separate CNTs by

length [113,122]. FFF is a chromatography-like separation

and sizing technique based on elution through a thin empty

channel. The main difference between FFF and chromatogra-

phy is that FFF separation is (ideally) induced only by physical

interactions with an external field rather than physicochem-

ical interactions with a stationary phase. Compared with

chromatography and CE, FFF can separate CNTs by length

over a larger range and in larger quantities.

These techniques can separate CNTs according to their

size or electronic properties, which represents an important

improvement in size and conductivity selectivity. This will

promote the development and application of CNTs in the ana-

lytical, nanotechnology and nanoelectronics fields [123,124].

Therefore, they are not merely purification methods. A com-

mon feature of these techniques is that they require high dis-

persibility of isolated CNTs in solution. However, the CNT

surface is hydrophobic, and the existing state of CNTs, espe-

cially for SWCNTs, is interconnected or in a thick bundle. In

order to obtain high-purity CNTs, pre-treatment is required

to obtain isolated CNTs having high dispersibility in solution.

2.2.7. Brief summary

Physical-based purification can maintain the intrinsic

structure of CNTs, which is desirable for elucidating their

properties. Furthermore, some techniques such as chroma-

tography, electrophoresis, and FFF can separate CNTs accord-

ing to their differences in length or conductivity in addition to

their purification function, which is a key step for using CNTs

in devices such as nano- and micro-electronics. However,

Fig. 9 AFM images of three representative chromatography fractions of SWCNTs deposited onto alkyl silane-coated SiO 2substrates (reprinted with permission from[111], Copyright 2005 Amercian Chemical Society).

C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5 2015


14/23

there are still some problems in these techniques that need to

be solved. One is that these methods are not very effective in

removing impurities. Another is that they require CNT sam-

ples be highly dispersible. Therefore, the as-prepared sample

is always first dispersed in solution by adding surfactants or

treated by a chemical process to cut and/or add functional

groups before purification. The third problem is the limited

amount of sample that can be purified each time. Based on

the above facts, physical methods are more suitable for use

as an assistant step combined with chemical purification, ex-

cept for the case where a small amount of CNTs with a partic-

ular structure or property are required.

2.3. Multi-step purification

As discussed above, gas phase oxidation is effective in remov-

ing amorphous carbon and polyhedral carbon at the cost of

losing some CNTs but fails to remove a significant fraction

of graphite particles and metal impurities. Liquid phase oxi-

dation with strong oxidants is effective in removing carbona-

ceous impurities and metal particles simultaneously, whereas

purified CNTs are always cut, opened and damaged. Physical

purifications are effective in partly removing isolated carbo-

naceous or metallic impurities, while amorphous and spher-

ical particles stuck to sidewalls or metal particles

encapsulated in CNTs remain. In order to achieve desirable

CNT purity with high yield, combinations of chemical and

physical purifications are being intensely investigated.

According to different needs, various kinds of multi-step puri-

fication methods are reported. For example, it is difficult to re-

move carbonaceous impurities adhering to the sidewalls of

CNTs for both chemical and physical purifications. To solve

this problem, hydrothermally initiated dynamic extraction

(HIDE) [44,125127] or sonication [128133] was adopted in

many chemical purification procedures. Graphite particles

existing in CNTs synthesized by arc discharge or laser abla-

tion are hard to remove by chemical oxidation, so filtration

is adopted in some purification[134,135]. It is clear that metal

particles catalyze the low-temperature oxidation of carbons

indiscriminately, destroying SWCNTs in the presence of oxy-

gen and other oxidizing gases. To overcome this problem,

purification combining gas phase oxidation and acid treat-

ment were widely investigated[136145]. In fact, several tech-

niques such as oxidation, sonication, HIDE, or filtration are

simultaneously adopted in one purification procedure to ob-

tain high-purity CNTs with high yield.

2.3.1. HIDE-assisted multi-step purification

HIDE provides comminution on a microscopic scale as a re-

sult of collision between soot particles and water molecules

during thermal treatment[146]. Thus during HIDE, water mol-

ecules break the network between SWCNTs, amorphous car-

bons and metal particles, and also attack the graphitic layers

encapsulating metal particles. As a result, almost all graphitic

nanoparticles and CNSs are washed out from the soot. The

graphitic sheets of the CNSs, for the most part, have defects

and dislocations, in contrast to SWCNTs [147]. It is believed

that the reaction of H2O with carbon breaks the graphitic lay-

ers that wrap the metal particles. Consequently, incorporat-

ing HIDE in the purification procedure exposes the metal

particles and this allows them to be dissolved by hydrochloric

acid in the final step of the treatment [125].

Tohji et al. [125,126]first reported a multi-step method to

purify SWCNTs by combining HIDE with other processes as

illustrated inFig. 10. SWCNTs with a purity of 95 wt% and a

yield of about 2 wt% were obtained by this process.

Graphite fragments from the graphite rod cannot be re-

moved by oxidation. To solve this problem, the as-prepared

MWCNTs were treated by a multi-step process (shown in

Fig. 11) combining wet grinding, HIDE, oxidation and other

techniques [127]. TEM observations indicated that this pro-

cess was effective in removing graphite and carbonaceous

particles and opening CNT caps. However, the yield was only

2 wt% due to the high oxidation temperature (700 C).

To increase the purification yield, we developed a multi-

step method combining sonication, HIDE, bromination, gas

phase oxidation and acid treatment [44]. It was found that

bromination can increase the purification yield from 25 to

50 wt%. The effect of bromination on the purification of

CNTs was also verified by Fan et al. [148]. The problem with

this technique is that it is not suitable for purifying SWCNTs.

On the other hand, the onset burning temperature of

MWCNTs was decreased after purification, suggesting that

defects or functional groups were introduced.

2.3.2. Microfiltration in combination with oxidation

Bandow et al. [134] purified SWCNTs synthesized by laser

ablation by combining microfiltration [86] with oxidation in

air. In a typical procedure, as-prepared soot containing

SWCNTs was first purified using microfiltration to remove

large CNSs. The obtained SWCNTs were then oxidized in air

at 450 C for 20 min to remove CNSs adhering to the SWCNT

walls, followed by soaking in concentrated HCl (36%) for 12

days at room temperature to remove metal particles. The pur-

ity of the SWCNTs after purification was greater than 90%.

To remove metal particles before oxidation, Kim and Luzzi

[135]developed magnetic filtration carried out in a magnetic

field. They investigated the efficiency of using magnetic filtra-

tion alone, or combining it with chemical-based or annealing-

assisted oxidation treatments. Using magnetic filtration

alone, the catalyst content was reduced from 11.7 to

3.7 wt%, much better than obtained in oxidation or chemical

As-produced

SWCNTs

HIDE treatment

for 12 h

Filtration

and dry

Soxhlet extraction

in CS2

Oxidation at 470C

for 20 min

Removing

fullerenes

Removing

amorphous carbon

Removing

exposed metal

6 M HCl

treatment

Purified

SWCNT

Fig. 10 A diagram showing the purification of SWCNTs

with a multi-step process incorporating HIDE treatment.

2016 C A R B O N 4 6 ( 2 0 0 8 ) 2 0 0 32 0 2 5


15/23

treated samples. By combining chemical and magnetic purifi-

cation, the metal catalyst content was reduced to 0.3 wt%.

These results allowed the authors to conclude that magnetic

filtration is effective in removing metal catalysts, producing

CNTs with high-quality and yield. It is well known that metal

particles can catalyze carbon oxidation in the presence of oxi-

dants. Therefore, effective removal of metal particles is desir-

able for the following purification of CNTs using chemical

oxidation. Thus magnetic filtration combined with chemical

purification opens a new way to obtain purified CNTs with

high yield.

2.3.3. Sonication in combination with oxidation

Sonication is identified as one of the effective processes to get

rid of the amorphous impurities adhering to the walls of

CNTs using suitable solvents[149]. During sonication, the sol-

vent molecules are able to interact with CNTs and hence lead

to solubilization, which can improve purification effective-

ness when some other steps are followed.

We[128]developed an effective multi-step purification ap-

proach to purify SWCNTs by combining sonication with oxi-

dation in air. The biggest problem in purifying SWCNTs

synthesized by arc discharge is how to remove graphite parti-

cles produced from the graphite rod. Ultrasonication in etha-

nol was adopted to first remove the graphite particles.

Because the SWCNTs used were rope-like, we decanted the

alcohol solution containing graphite particles and other or-

ganic impurities after sonication for about 5 min. This proce-

dure was repeated five times. The graphite-free material was

oxidized and then soaked in HCl to remove amorphous

carbon and metal particles. TEM observations (Fig. 12) and Ra-

man spectra verified the effectiveness of the above purifica-

tion procedure. With this procedure, a 41 wt% yield of

SWCNTs with a purity of about 96% was achieved. However,

the above procedure is only applicable to SWCNTs synthe-

sized by hydrogen arc discharge, in which SWCNTs exist as

ropes with fewer defects.

We [129] also developed a multi-step method to purify

SWCNTs synthesized by CVD, which includes acid washing,

ultrasonication and freezing treatments in liquid nitrogen.

After purification, SWCNTs with a purity of 95% (estimated

from EDS and SEM) and a yield of 40 wt% were obtained and

the procedure did not destroy the SWCNT bundles.

Montoro and Rosolen[130]reported a four-step method to

purify SWCNTs synthesized by arc discharge as shown in

Fig. 13. This new procedure is efficient and appropriate for

obtaining highly pure SWCNTs with minimum damage to

the CNT walls and minimum modification in the CNT length.

Wang et al.[132]developed a three-step method to purify

and cut SWCNTs synthesized by CVD. This method included

refluxing in 2.6 M HNO3to remove metal particles, ultrasoni-

cation in acid solution (H2SO4/HNO3, H2SO4/H2O2) to cut and

polish the SWCNTs, and heat treatment in an NH3 atmo-

sphere to remove carbon impurities and heal structural de-

fects. Recently, they [133] further impro

Documents

713 Puriﬁcation of Carbon Nanotubes