FULL PAPER
Single-Walled Carbon Nanotubes
Tuning Geometry of SWCNTs by CO2 in Floating Catalyst
CVD for High-Performance Transparent Conductive Films
Yongping Liao, Aqeel Hussain, Patrik Laiho, Qiang Zhang,* Ying Tian, Nan Wei,
Er-Xiong Ding, Sabbir A. Khan, Nguyen Ngan Nguyen, Saeed Ahmad,
and Esko I. Kauppinen*
Optimized geometry of single-walled carbon nanotubes (SWCNTs) is vital to
high-performance transparent conductive films (TCFs). Herein, the geom-
etry of SWCNTs, i.e., tube diameter, bundle length, and bundle diameter, are
successfully tuned by introducing carbon dioxide (CO2) into floating catalyst
chemical vapor deposition (FC-CVD), where carbon monoxide (CO) is used
as a carbon source and ferrocene as a catalyst precursor. Both tube diameter
and bundle length increase with an increment of CO2 concentration, and the
yield of SWCNTs can be significantly promoted with the appropriate amount
of CO2. The role of CO2 in this behavior is further rationalized as enhancing
CO decomposition or carbon dissolution into catalysts. The TCFs based on
SWCNTs with optimized geometry by CO2 exhibit improved performance up
to 86.8 Ω sq−1 at 90% transmittance after AuCl3 doping, achieving about 50%
reduction of sheet resistance compared to the TCFs without CO2. The use of
CO2 for directly tuning growth of SWCNTs blazes a new trail in the field of
SWCNT based TCFs.
1. Introduction
Transparent conductive films (TCFs) have attracted great atten-
tion because of their wide applications, such as solar cells,[1,2]
light-emitting diodes,[3] and touch sensors.[4,5] Generally, the
materials used for TCFs should have electrical charge transport
and optical transmittance merits. The most widely used mate-
rial for TCFs is indium–tin oxide (ITO). However, ITO is poorly
suited for flexible displays, limiting its further application.
Therefore, it is necessary to explore new materials to replace it.
In recent decades, the emergence of alternatives like graphene,
Y. P. Liao, A. Hussain, Dr. P. Laiho, Dr. Q. Zhang, Prof. Y. Tian,
Dr. N. Wei, E.-X. Ding, S. A. Khan, N. N. Nguyen, S. Ahmad,
Prof. E. I. Kauppinen
Department of Applied Physics
Aalto University School of Science
P.O. Box 15100, FI-00076 Aalto, Finland
E-mail: qiang.zhang@aalto.fi; esko.kauppinen@aalto.fi
Prof. Y. Tian
Department of Physics
Dalian Maritime University
Dalian, Liaoning 116026, China
The ORCID identification number(s) for the author(s) of this article
can be found under https://doi.org/10.1002/admi.201801209.
DOI: 10.1002/admi.201801209
Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (1 of 10)
www.advmatinterfaces.de
silver nanowire, and single-walled carbon
nanotubes (SWCNTs) brings new oppor-
tunities to TCFs’ research.[6–11] Among
these alternatives, while popular, graphene
normally has grain boundaries or a poor
interlayer junction contact, which limits
charge carrier transport in TCFs.[12] Due to
its cost and instability, silver nanowire is
also restricted to scale-up applications.[13]
In contrast, TCFs based on SWCNTs have
unique electrical conductivity and excel-
lent mechanical properties that enable the
production of flexible electronics at low
cost.[14,15]
To ensure the high performance of
SWCNT-based TCFs, researchers have
attempted different methods to improve
SWCNT film conductivity, for instance,
by doping SWCNTs with acid or metal
oxide,[16,17] fabricating hybrids of SWCNTs
with graphene or polymers,[18–20] or
enriching the ratio of metallic SWCNTs.[21,22] Although these
methods improve the conductivity of SWCNT films, some
issues still remain, such as the tube damage and impurities
caused by the solution process, other tedious experimental pro-
cesses, and high overall costs. Nevertheless, it is possible to pre-
serve the intrinsic properties of SWCNTs using floating catalyst
chemical vapor deposition (FC-CVD).
It is well known that the electrical conductivity of SWCNT
TCFs is mainly determined by contact resistance of junc-
tions within the network.[23] The diameters of the nanotubes
and bundles have a significant effect on junction resistance,
and the bundle length can determine the number of contact
junctions. Our previous work indicates that SWCNTs with
a larger diameter show lower contact resistance than those
with a smaller diameter.[24] Obrzut and co-workers and Jeong
and co-workers study the effect of SWCNT length on conduc-
tivity and optical properties.[25,26] Liu and co-workers stress
the importance of the geometry of double-walled carbon
nanotubes when improving conductivity.[27] Our previous
work also found that less-bundled SWCNTs show excel-
lent conductivity.[28] Therefore, a controllable synthesis of
SWCNTs with a relatively large tube diameter, long length,
and less bundling is an efficient approach to improve the
conductivity of TCFs.
One possible approach is introducing certain amount
of CO2 to the reactor. CO2 has been reported to tune CNT
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
www.advmatinterfaces.de

growth when hydrogen carbon, such as CH4 and C2H4, was
used as a carbon source. Researchers find that CO2 can cause
significant effects not only on the diameter[29] and the wall
number,[30] but also on the yield of the CNTs.[31–34] Never-
theless, few works were reported about CO2 effect on CNT
growth when CO is used as a carbon source. In our previous
work, a sheet resistance of 200 Ω sq−1 at 90% transmittance
after doping is achieved using CO as the carbon source (but
without CO2).[14] In that case, employing CO2 might have
more prominent effect on SWCNT growth due to the Bou-
douard reaction (2CO ↔ C + CO2); CO2 may affect not only
the CO decomposition but also catalyst activity. Furthermore,
there is no mechanism that can fully explain the role of CO2,
and no thin films are studied using CO2. Therefore, it is
promising to investigate the role of CO2 and achieve high-
performance SWCNT TCFs using CO2 with CO as carbon
sources.
In this work, various amounts of CO2 were introduced to
tune the growth of SWCNTs in FC-CVD, in which CO was
used as a carbon source and ferrocene as a catalyst. The
results demonstrated that in addition to the geometry of
SWCNTs, CO2 could also have a clear effect on the yield of the
process and the catalytic particles. Under an optimized CO2
volumetric concentration of 0.50%, we obtained SWCNTs
with a relatively large diameter, long length, and less bun-
dling. The corresponding TCFs showed an excellent perfor-
mance of 86.8 Ω sq−1 under 90% transmittance at 550 nm
(T550  nm) after AuCl3 doping.
2. Results and Discussion
2.1. Effect of CO2 on the Geometry of SWCNTs
2.1.1. Effect of CO2 on SWCNT Diameter
UV–vis–NIR spectroscopy is an efficient and comprehensive
method to analyze the diameter distribution, since all tubes will
contribute to the absorption spectra. As shown in Figure 1a,
SWCNT thin films synthesized at CO2 concentration from 0%
to 0.74% exhibit the three absorption peaks from the first and
second electronic transitions of semiconducting nanotubes (S11
and S22) and the first electronic transition of metallic nanotubes
(M11), respectively. With the increase of CO2 concentration,
these absorption peaks gradually shift to longer wavelengths,
and such a shift indicates an increase in the diameter of the
SWCNTs. By using the fitting method based on absorption
spectra,[35] we obtained the mean diameter of SWCNT samples
increased from 1.1 nm at 0% CO2 to 2.1 nm at 0.74% CO2.
The fitted diameter distribution of three samples is shown in
Figure 1b–d. With further increasing CO2 concentration up to
1.23%, we observe a rather flat absorption profile, suggesting
low SWCNT fraction in this sample. This result is also observed
in the scanning electron microscopy (SEM) and transmission
electron microscopy (TEM) images (Figure S1, Supporting
Information). Remarkably, the increase of CO2 concentration
accounts for SWCNT diameter enlargement. Considering that
such a trend is consistent with our earlier research that was

Figure 1.  The mean diameter of SWCNTs affected by CO2 based on absorption spectra. a) Absorption spectra of prepared SWCNT films with different
CO2 concentrations; the inset shows the mean diameter of SWCNTs with various CO2 concentrations. Fitted diameter distribution for SWCNTs with
CO2 concentration of b) 0%, c) 0.31%, and d) 0.50%, respectively. The mean diameter of SWCNTs increases from 1.1 to 2.1 nm when increasing CO2
concentration from 0% to 0.74%.

Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (2 of 10) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
Figure 2.  The diameter and quality of SWCNTs affected by CO2 from Raman spectra. RBM bands with excitation laser wavelength of a) 488, b) 514,
and c) 633 nm. The RBM peaks shift to the smaller wavenumber range with increase of CO2 concentration, indicating the enlargement of SWCNT
diameter. G and D bands excited by d) 633 nm laser; the IG/ID increases with CO2 concentration, stating that the SWCNT quality is improved by CO2.
done at 880 °C and 300 ccm CO flow rate through ferrocene,[29]
the influence of CO2 on the nanotube diameter distribution is
considered a general phenomenon. In addition, by optimizing
the CO2 concentration introduced to the reaction, both the
quality and yield of SWCNTs in this work have been promoted.
Raman spectra at excitation wavelength of 488, 514, and
633 nm were used to further characterize the properties of
SWCNTs. As shown in Figure 2a–c, the Raman spectra of the three
SWCNT thin films exhibit the characteristic radial breathing
mode (RBM). The diameter of SWCNTs can be calculated from
the RBM frequency (ω) with ω = 234/d + 10.[36] For instance, with
the excitation of a 514 nm laser (Figure 2b), the RBM peaks for
SWCNTs with 0% CO2 are mainly in the range of 175–200 cm−1.
Increasing the CO2 concentration from 0% to 0.50% shifts the
RBM peaks to a lower frequencies, indicating the diameter of the
tubes becomes larger. At 0.50%, the RBM peaks are in the range
of 100–150 cm−1, suggesting this sample has the largest diameter
among the three. The similar trends are also found under the
other two excitation wavelengths of 488 and 633 nm (Figure 2a,c).
The diameter analysis from Raman spectra is consistent with
that from absorption spectra. In addition, the intensity ratio of
G (cm−1) and D (cm−1) band (IG/ID) can reflect the quality of
SWCNTs.[37] As shown in Figure 2d, the IG/ID values of SWCNTs
with CO2 are higher than the one without CO2. This means the
quality of SWCNTs has been improved with the presence of CO2.
A previous study indicated that a proper amount of CO2 can etch
amorphous carbon, resulting in the increase of IG/ID values.[38]
However, an excessive amount of CO2 will degrade the quality of
SWCNTs (Figures S1 and S2, Supporting Information).
2.1.2. CO2 Increases the Bundle Length
As shown in Figure 3a, the bundle length for SWCNTs synthe-
sized at 0% CO2 is very short, of which the mean length is of
Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (3 of 10)
www.advmatinterfaces.de
2.7 µm on the basis of statistical calculation from 305 bundles
(Figure 3d). After introducing 0.31% of CO2 to a reactor, the
bundle length increases significantly, around 4.1 µm (307 bun-
dles, Figure 3b,e). It even elongates to 7.5 µm when 0.50%
CO2 is fed (299 bundles, Figure 3c,f). Interestingly, the long
bundles are still straight (Figure 3c), implying better electrical
transport. However, higher CO2 concentration of 1.23% did
not lengthen the bundles but keep them at the mean length of
only 7.4 µm (Figure S3, Supporting Information). This is due to
the suppressed CO decomposition by excessive CO2, leading to
insufficient carbon feeding. Therefore, the continuous growth
in length is suppressed. Consequently, among the three sam-
ples (Figure 3a–c), SWCNTs synthesized at 0.50% CO2 imply
the least number of contact junctions (under same transmit-
tance) because of the long length, which could increase the
electrical conductivity of SWCNT TCFs.
2.1.3. Bundle Diameter Affected by CO2
The bundle diameter of SWCNTs is also crucial for electrical
conductivity.[39] Here, the mean bundle diameter of SWCNTs
synthesized at different CO2 concentrations was averaged
from 200 bundles for each sample using TEM (Figure 4b–g).
According to the TEM results, the as-prepared SWCNTs are
clean with little amorphous carbon (Figure 4b–d). The relatively
small mean bundle diameter of 7.5 nm at 0% CO2 (Figure 4b,e)
increases up to 9.5 nm at 0.31% CO2 (Figure 4c,f). However,
it reduces to 7.1 nm as CO2 concentration increases to 0.50%
(Figure 4d,g). We then attempt to rationalize this trend of
bundle diameter by correlating the bundle formation and the
yield of SWCNTs.
Considering that higher yield results in the higher possi-
bility of collision between SWCNTs in the aerosol phase, we
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
www.advmatinterfaces.de

Figure 3.  Bundle length with various CO2 concentration. Typical SEM images of bundle length and corresponding length statistics with various CO2
concentration of a,d) 0%, b,e) 0.31%, and c,f) 0.50%. The bundle length is elongated by employing CO2. Specifically, the bundle length with 0.50%
CO2 is increased to 7.5 ± 5.6 µm, showing great potential for improved TCF conductivity.

Figure 4.  The yield and bundle diameter affected by CO2. a) Transmittance after collecting for 20 min and the normalized yield of SWCNT films with
various CO2 concentrations. Typical TEM images of SWCNT bundle and corresponding statistics of bundle diameter distribution with CO2 concen-
tration of b,e) 0%, c,f) 0.31%, and d,g) 0.50%. The yield is promoted significantly by certain amount of CO2 concentration, for instance, the yield of
SWCNTs with 0.31% CO2 is much higher than those with 0% CO2. Since the higher the yield, the higher possibility the tubes collide in the aerosol
phase, leading to the larger bundle diameter. So the sample with 0.31% CO2 has the largest bundle diameter among these three samples. SWCNTs
with 0.50% CO2 have relatively small bundle diameter, which is beneficial for better TCF conductivity.

Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (4 of 10) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
www.advmatinterfaces.de

are expecting a strong relationship between yield and bundle
diameter. To quantify the yield, we use the Beer–Lambert law,
A  =  −lnT  = ερL (A is the absorbance, T is the transmittance,
ε is the extinction coefficient, ρ is the density of film, and L is
the thickness of film), then we can get L = −lnT/ερ. Since L is
proportional to the yield, we can obtain the information of yield
based on optical transmittance. For instance, we measured the
transmittance of SWCNT films from different CO2 concentra-
tions within the same collection time (20 min); based on the
transmittance, the normalized yield with various CO2 con-
centration can be obtained (Figure 4a). The filter used here to
collect SWCNT films was 1.33 cm2. As depicted in Figure 4a,
the yield is enhanced significantly with the proper amount of
CO2. It is over two times higher with 0.37% CO2 than the one
without CO2. However, the yield started to decrease when the
CO2 concentration was further increased to 0.50%. This pheno­
menon can be explained by the Boudouard reaction
2CO(g) ↔ C(s) + CO2(g) , ∆H = −169 kJ mol −1 (1)
Notably, without adding extra CO2 into the reactor, CO2
concentration from the outlet of the reactor is quite low due
to the low disproportionation rate of CO.[38] So even adding a
little amount of CO2 can affect the reaction. However, too much
CO2, such as 1.23%, will suppress the CO decomposition by
shifting the Boudouard reaction backward, so the yield as well
as the quality of nanotubes decreases significantly. Hence, the
trend of bundle diameter well matches with that of the yield,
further proving the relationship between bundle and yield, that
is, higher yield will lead to larger bundle diameter.
2.2. Catalysts Affected by CO2
It is well known that catalysts correlate to SWCNT growth
closely. Thus, to investigate how the catalysts influence the
growth of SWCNTs, TEM was applied to analyze the size and
chemical composition of catalysts, and differential mobility
analysis (DMA) was further utilized to monitor the number
size distribution of catalysts and SWCNTs in gas phase from
the outlet of reactor. After careful analysis of catalyst size (about
200 catalyst particles for each sample) with various CO2 con-
centrations from TEM images (Figure 5a–c), we found that
regardless of the introduced CO2 concentration, the size distri-
bution of catalysts is almost similar with a mean size of 2.7 nm
(Figure 5d–f). It implies that the addition of CO2 did not alter
the diameter of catalysts, but the diameter of SWCNTs can
be different from the same size of catalyst. Furthermore, the
chemical composition of catalysts is also important for SWCNT
growth. We found that the chemical composition of catalysts
has no difference with various CO2 concentrations (Figure S4,
Supporting Information). Figure 5g shows that the total con-
centration of SWCNTs increases with a certain amount of CO2
such as 0.31%, indicating that the number of SWCNTs and
the activated catalysts increases with a certain amount of CO2.
The trend of yield here is consistent with what we measured in
Figure 4a.
2.3. Schematic of SWCNT Growth Modulated by CO2
Based on an analysis of SWCNT geometry, yield, and catalysts
influenced by CO2, we propose a schematic illustrating how
CO2 affects SWCNT synthesis (Figure 6). Theoretically, in the
aerosol phase, particles can deposit on the quartz wall sur-
face by diffusion;[40] the deposition ratio depends on the size
of aerosol particles at a given temperature, quartz tube length,
and flow rate. In our experimental conditions, ≈75% of nano-
particles will deposit on the quartz wall when there are only
nanoparticles with the diameter of 2.7 nm in the gas phase
(see the calculation details in the Supporting Information).
However, as soon as the nanoparticles start to quickly catalyze

Figure 5.  Catalyst size with various CO2 concentrations. Representative TEM images for analyzing catalysts with CO2 concentration of a) 0.50%,
b) 0.31%, c) 0%, and d–f) the corresponding catalyst size distribution, respectively. It turns out that the addition of CO2 cannot change the catalyst
size, but the diameter of SWCNTs grown on catalyst varies with CO2 concentration. g) Total concentration from outlet of reactor with various CO2
concentrations measured by DMA, indicating that the number of SWCNTs and activated catalysts increases with 0.31% CO2, but then decreases with
0.50% and 1.23% CO2.

Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (5 of 10) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
Figure 6.  A schematic illustrating the SWCNT growth with 0% (left), 0.31% (middle), and 0.50% (right) of CO2, (the SWCNT bundle is not displayed here).
the nucleation and growth of carbon nanotubes, the deposition
ratio will decrease due to the increase of mobility diameter.
Furthermore, the carbon coating of catalysts could suppress
the carbon source decomposition on a catalyst surface and
shorten the lifetime of catalyst. As demonstrated in Figure 6,
introducing a certain amount of CO2 could probably enhance
CO decomposition or remove amorphous carbon on catalysts,
depending on the CO2 concentration.
In detail, when no extra CO2 is introduced (Figure 6, left),
the CO decomposition is quite low. The low decomposition
leads to insufficient carbon feeding to the catalysts so that the
high amount of catalysts deposit on the quartz wall by diffu-
sion, and finally results in low yield and short tubes.
When 0.31% CO2 is employed into the reactor (Figure 6,
middle), the dominant function of CO2 is to remove the amor-
phous carbon coated on a catalyst surface, enhancing the
decomposition of carbon monoxide to carbon by shifting
the Boudouard reaction forward. Thus, more catalysts will be
activated for growing SWCNTs due to the enhanced carbon
feeding, and fewer catalysts will deposit on the wall, leading to
higher yield and longer tubes.
When CO2 concentration increases slightly to 0.50%
(Figure 6, right), the Boudouard reaction will shift backward
due to the slight excess of CO2. Hence, the CO decomposition
is reduced modestly and more catalyst will deposit on the wall,
leading to lower yield. On the other hand, the excessive CO2
can remove coated carbon on a catalyst. Therefore, the carbon
dissolution and diffusion in catalysts is promoted, leading to a
faster growth rate of SWCNTs. Finally, much longer tubes are
consequently obtained.
However, with a highly excessive 1.23% CO2, the decompo-
sition of CO is heavily suppressed, resulting in much lower
Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (6 of 10)
www.advmatinterfaces.de
carbon feeding into catalysts, and some SWCNTs might be
etched away with a high concentration of CO2. Thus, the small
nanoparticles which are not covered by carbon will collide and
aggregate to larger clusters, finally, flowing out of the reactor.
This assumption well agrees with what we observed from SEM
and TEM (Figure S1, Supporting Information).
2.4. SWCNTs with Various Geometry for TCFs
Based on the above analysis and literature results, we assume
that the TCFs based on SWCNTs with a large tube diameter,
long length, and less bundling could potentially exhibit
improved conductivity. Therefore, SWCNT TCFs with var-
ious transmittance were fabricated via a dry transfer method.
The plot of transmittance versus sheet resistance was fitted
according to the relationship between sheet resistance (Rs) and
transmittance (T) in Equation (2)
−1
Rs = (2)
KlnT
where K is a coefficient of proportionality, which contains both
optical and electrical characteristics of the film and can be used
as a figure of merit to characterize the quality of film.[14]
The pristine TCFs with different CO2 concentration exhibit
various levels of conductivity (Figure 7a). On one hand, the
sheet resistance of SWCNT films at 90% T550nm decreases
from 1075.6 Ω sq−1 at 0% CO2 to 273.4 Ω sq−1 at 0.50% CO2.
On the other hand, SWCNT films synthesized at 1.23% CO2
show inferior conductivity due to the excess nanoparticles
(Figures S1 and S2, Supporting Information). To further
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
www.advmatinterfaces.de

Figure 7.  Performance of SWCNT TCFs with various CO2 concentrations. Sheet resistance versus transmittance at 550 nm of SWCNT TCFs a) before
and b) after doping with AuCl3, the inset shows the sheet resistance at 90% T550  nm for SWCNT films with different CO2 concentrations. The SWCNT
TCF with 0.50% CO2 displays a sheet resistance of 86.8 Ω sq−1 at 90% T550  nm after doping, which is the lowest among all the samples in this work.
The improved conductivity of SWCNT TCF mainly results from the optimized geometry by CO2. c) Sheet resistance ratio of pristine to doped films with
various CO2 concentrations at different transmittance. d) Doping ratio of SWCNT films with various CO2 concentrations calculated from panel (c).
e) Typical SEM image of SWCNT film with CO2 concentration of 0.50% and f) optical transmittance as a function of light wavelength for pristine SWCNT
film on PET; the inset is the corresponding film used as conductive channel to light up a blue LED.

improve the conductivity of SWCNT TCFs, AuCl3 was used
as an efficient dopant.[41] The absorption spectra before and
after doping are displayed in Figure S5 (Supporting Informa-
tion). While doping by AuCl3 boosts up the conductivity of all
SWCNT TCFs, the trend is not different (Figure 7b). In fact,
our best sample is the doped SWCNT film synthesized at 0.50%
CO2 with a sheet resistance as low as 86.8 Ω sq−1.
Contact junctions play a crucial role in the conduc-
tivity of SWCNT TCFs. As demonstrated above, the conductivity
of SWCNT TCFs can be improved by decreasing the number
of contact junctions as well as the junction resistance. As com-
pared to the growth without CO2, SWCNT films synthesized
at 0.50% of CO2 have two advantages: i) less number of junc-
tions due to the longer tubes and ii) lower junction resistance
due to the larger tube diameter and less bundling. Known from
Figures 1 and 4, compared with the 0% CO2 sample, the 0.50%
CO2 sample has a larger tube diameter but a smaller bundle
diameter; thus, the number of tubes in one bundle will be
decreased, and more tubes can join in the conductive network
to improve the conductivity. Thus, SWCNT films with 0.50%
CO2 exhibit the best conductivity among all samples. This is
clear evidence of our assumption that finely tuning the geom-
etry of SWCNTs can improve the conductivity of TCFs.
According to the sheet resistance of SWCNT films before and
after doping (Figure 7a,b), we calculated the sheet resistance
ratio of pristine to doped films (doping factor) with various CO2
concentrations at different transmittance values (Figure 7c,d).
We can see that with a given CO2 concentration, SWCNT films
undergo similar doping factor at different transmittance values
(Figure 7c). On the other hand, the doping factors vary with
CO2 concentration. As shown in Figure 7d, SWCNT films with
0% CO2 display a mean doping factor of ≈6; then, it decreases
to about 5.4 and 3.1 with 0.31% and 0.50% CO2, respec-
tively. Yet, such a factor rises to ≈3.8 and 5.4 with higher CO2

Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (7 of 10) © 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
Figure 8.  A comparison of our SWCNT TCFs to other reported results. Our
strategy of using CO2 can achieve higher-performance SWCNT TCFs than
many of the reported results. Compare to the reported methods, our method
of employing CO2 is easier to operate and the yield is relatively high.
concentration of 0.62% and 0.74%, respectively. This pheno­
menon can be explained by the fact that chemical doping
mainly lowered the contact resistance between SWCNTs.[24]
Herein, compared to SWCNTs with 0% and 0.31% CO2,
SWCNTs with 0.50% CO2 possess fewer contact junctions, as
well as lower junction resistance. Additionally, through fur-
ther analysis of chirality of our tubes by electron diffraction,
we found that the metallic tube ratio could be increased with
the increment of CO2 concentration. SWCNTs with 0.50% CO2
have higher metallic tubes ratio than those with 0% and 0.31%
CO2,[42] lowering the probability of metallic–semiconducting
contact. Therefore, the junction resistance of pristine film with
0.50% CO2 is lower than with 0% and 0.31%, and the dopant
will have less effect on SWCNTs with 0.50% than with 0% and
0.31%. Consequently, SWCNT films with 0.50% CO2 exhibit
a lower doping factor. However, with too much CO2 such as
0.62% and 0.74%, SWCNTs may have more amorphous carbon
on the nanotube walls. Such amorphous carbon can increase
the junction resistance in the pristine films, which makes the
doping factor higher.
The SWCNT bundles can form an arbitrary network, as
shown in Figure 7e. The clean SWCNT network poses good
interconnection between tubes, which is also an important
factor for improving its conductivity. As demonstrated in
Figure 7f, a SWCNT film (0.50% CO2) on polyethylene tere-
phthalate (PET) with 90% T550nm was used as the conductive
channel to light up a blue light-emitting diode (LED).
A comparison of our SWCNT TCFs to other
research[4,14,24,39,43–47] reported recently is shown in Figure 8.
This suggests that our SWCNT TCFs performed better than
most other approaches except those reported by Mustonen
et al.,[39] Kaskela et al.,[43] and Hecht et al.[45] However, in
Mustonen et al.’s work, the yield of SWCNT TCFs was much
lower than this work,[39] which limits real applications. Fur-
thermore, Kaskela et al. used the patterned SWCNT films to
increase the conductivity.[43] Hecht et al. introduced a super acid
to improve the optoelectronic performance.[45] On the other
hand, our films exhibit even lower sheet resistance than the
patterned films reported by Fukaya et al.[4] In short, using CO2
Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (8 of 10)
www.advmatinterfaces.de
in this work is a facile method to achieve high-performance
SWCNT TCFs with relatively high yield.
3. Conclusion
In summary, we have successfully tuned the geometry of
SWCNTs by controlling CO2 concentration in a FC-CVD system
and further investigated the performance of TCFs based on
SWCNTs with various geometries. The CO2 concentration intro-
duced to the reactor is a key factor in obtaining various geom-
etries of SWCNTs. We found that raising CO2 concentration
(from 0% to 0.74%) increases both the tube diameter (from 1.1
to 2.1 nm) and length (from 2.7 to 7.5 µm). Additionally, the
yield can also be boosted over two times with a proper amount
of CO2 (0.31%). Based on the effect of CO2 on SWCNT growth,
we propose that the role of CO2 is to enhance the dispropor-
tionation of CO or carbon dissolution and diffusion in the cata-
lyst, depending on CO2 concentration. TCFs based on SWCNTs
with optimized tube diameter (1.9 nm), length (7.5 µm), and
bundle diameter (7.1 nm) showed an excellent sheet resistance
of 86.8 Ω sq−1 at 90% T550  nm. The enhanced conductivity of
SWCNT TCFs mainly results from the reduced contact junc-
tions between SWCNTs. In this work, the direct use of CO2
offers promising approaches to in situ control of the SWCNT
synthesis and provides great potential to scale-up production of
high-performance SWCNT TCFs.
4. Experimental Section
Materials and Reactor: Ferrocene (99.0%) was purchased from Alfa
Aesar, and carbon monoxide (99.0%) and carbon dioxide (99.9993%)
were obtained from AGA Industrial Gases. SWCNTs were synthesized
in house based on the thermal decomposition of ferrocene under
an atmosphere of CO in an FC-CVD system;[14] see Figure S6 in the
Supporting Information for a schematic of the FC-CVD reactor.
Synthesis of SWCNTs: A flow of 50 ccm CO was passed through the
cartridge at ambient temperature, where the ferrocene mixed with silica
sand was stored. The ferrocene-containing gas was injected through
a water-cooled injector probe maintained at a constant temperature
of 24 °C. The heating length of the reactor was 87 cm and the inner
diameter of quartz tube was 22 mm. Another pure CO flow of 250 ccm
was introduced from the main inlet and 100 ccm from bypass inlet.
The ferrocene vapor immediately decomposed into iron vapor after
coming out of the injector probe, followed by the nucleation to iron
nanoparticles in the quartz tube with a maximum temperature of 850 °C.
Subsequently, the iron nanoparticles catalyzed the growth of SWCNTs
inside the reactor with a laminar flow. DMA was used to determine the
number size distributions in the gas phase. In this work, a system of
scanning mobility particles sizer plus an electrometer (SMPS + E, Grimm
Aerosol Technic) was applied to monitor the number size distributions
from the outlet of the reactor. The prepared SWCNTs were collected
downstream of the reactor using a membrane filter with the help of a
vacuum. CO2 was used to tune growth of SWCNTs. In this work, extra
introduction of CO2 flow rates was set to 0.0, 1.0, 1.25, 1.5, 2.0, 2.5,
3.0, and 5.0 ccm corresponding to the volumetric fractions of 0%,
0.25%, 0.31%, 0.37%, 0.50%, 0.62%, 0.74%, and 1.23%, respectively.
For subsequent absorption, sheet resistance, or Raman measurement,
the SWCNT films were directly transferred onto quartz or PET substrates
without any purification.
Doping of SWCNTs: To increase the electrical conductivity of SWCNT
films, they were doped with AuCl3 according to previous work.[41]
Specifically, AuCl3 was completely dissolved in acetonitrile with a
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
concentration of 0.016 mol L−1, then the AuCl3 solution was dropped on
the SWCNT film. After the SWCNT film was dried in ambient air, more
acetonitrile was used to rinse the excess AuCl3, then the SWCNT film
was dried with compressed air. Finally, the sheet resistance after doping
could be measured.
Characterization: To measure the length of nanotubes, a
thermophoresis-based system was used to deposit SWCNTs on a SiO2/
Si substrate for 30 s,[48] followed by SEM imaging measurement. As
SEM could not accurately resolve whether the nanotube on SiO2/Si was
individual or a bundle, thus, the bundle length was counted using ImageJ
software from all the fractions observed in SEM. To collect samples for
TEM measurement, the TEM grid was put on a filter to harvest SWCNTs
for 20 s.
The morphology of SWCNT networks and the bundle length of
SWCNTs were characterized by SEM (ZEISS Sigma VP) under 1.0 kV, and
the bundle diameter of SWCNTs and nanoparticle size were measured
by high-resolution TEM (HR-TEM, JEOL JEM-2200FS) under 200 kV.
Raman spectroscopy (Horiba LabRAM HR 800) was used to estimate
the diameter and quality of SWCNTs with three excitation lasers of 488,
514, and 633 nm wavelengths. Absorption spectra were obtained with a
Perkin-Elmer Lambda 950 UV–vis–NIR spectrometer with wavelengths
ranging from 175 to 3300 nm. Sheet resistances were measured by a
Hewlett Packard 3485A multimeter 4-point probe system.
Supporting Information
Supporting Information is available from the Wiley Online Library or
from the author.
Acknowledgements
The authors acknowledge Dr. Hua Jiang’s help with TEM experiments
and refining of manuscripts. This work was supported by funding from
the European Union Seventh Framework Programme (FP7/2007-2013)
under Grant Agreement No. 604472 (IRENA project) and the Aalto
Energy Efficiency (AEF) Research Program through the MOPPI project.
The research was also partially supported by the Academy of Finland
via projects 286546 (DEMEC) and 292600 (SUPER) as well as by
TEKES Finland via projects 3303/31/2015 (CNT-PV) and 1882/31/2016
(FEDOC). This work made use of the Aalto University Nanomicroscopy
Center (Aalto-NMC) premises.
Conflict of Interest
The authors declare no conflict of interest.
Keywords
CO2, geometry, single-walled carbon nanotubes, transparent conductive
films
Received: August 6, 2018
Revised: September 4, 2018
Published online: October 4, 2018
[1] K. Aitola, K. Sveinbjörnsson, J.-P. Correa-Baena, A. Kaskela,
A. Abate, Y. Tian, E. M. J. Johansson, M. Grätzel, E. I. Kauppinen,
A. Hagfeldt, G. Boschloo, Energy Environ. Sci. 2016, 9, 461.
[2] I. Jeon, K. Cui, T. Chiba, A. Anisimov, A. G. Nasibulin,
E. I. Kauppinen, S. Maruyama, Y. Matsuo, J. Am. Chem. Soc. 2015,
137, 7982.
Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (9 of 10)
www.advmatinterfaces.de
[3] J. Song, P.-X Hou, M.-L. Chen, B.-W. Wang, D.-M. Sun,
D.-M. Tang, Q. Jin, Q.-X. Guo, D.-D. Zhang, J.-H. Du, K.-P. Tai,
J. Tan, E. I. Kauppinen, C. Liu, H.-M. Cheng, Sci. Adv. 2018, 4,
eaap9264.
[4] N. Fukaya, D. Y. Kim, S. Kishimoto, S. Noda, Y. Ohno, ACS Nano
2014, 8, 3285.
[5] W. Lee, H. Koo, J. Sun, J. Noh, K.-S. Kwon, C. Yeom, Y. Choi,
K. Chen, A. Javey, G. Cho, Sci. Rep. 2015, 5, 17707.
[6] X. Li, G. Zhang, X. Bai, X. Sun, X. Wang, E. Wang, H. Dai, Nat.
Nanotechnol. 2008, 3, 538.
[7] J. Zhao, S. Pei, W. Ren, L. Gao, H.-M. Cheng, ACS Nano 2010, 4,
5245.
[8] X. Li, Y. Zhu, W. Cai, M. Borysiak, B. Han, D. Chen, R. D. Piner,
L. Colombo, R. S. Ruoff, Nano Lett. 2009, 9, 4359.
[9] L. Hu, H. S. Kim, J.-Y. Lee, P. Peumans, Y. Cui, ACS Nano 2010, 4,
2955.
[10] S. De, T. M. Higgins, P. E. Lyons, E. M. Doherty, P. N. Nirmalraj,
W. J. Blau, J. J. Boland, J. N. Coleman, ACS Nano 2009, 3,
1767.
[11] Z. Wu, Z. Chen, X. Du, J. M. Logan, J. Sippel, M. Nikolou,
K. Kamaras, J. R. Reynolds, D. B. Tanner, A. F. Hebard, Science
2004, 305, 1273.
[12] K. S. Kim, Y. Zhao, H. Jang, S. Y. Lee, J. M. Kim, K. S. Kim, J. H. Ahn,
P. Kim, J. Y. Choi, B. H. Hong, Nature 2009, 457, 706.
[13] D. Langley, G. Giusti, C. Mayousse, C. Celle, D. Bellet,
J. P. Simonato, Nanotechnology 2013, 24, 452001.
[14] A. Kaskela, A. G. Nasibulin, M. Y. Timmermans, B. Aitchison,
A. Papadimitratos, Y. Tian, Z. Zhu, H. Jiang, D. P. Brown,
A. Zakhidov, E. I. Kauppinen, Nano Lett. 2010, 10, 4349.
[15] S. Park, M. Vosguerichian, Z. Bao, Nanoscale 2013, 5,
1727.
[16] H.-Z. Geng, K. K. Kim, K. P. So, Y. S. Lee, Y. Chang, Y. H. Lee, J. Am.
Chem. Soc. 2007, 129, 7758.
[17] S. L. Hellstrom, M. Vosgueritchian, R. M. Stoltenberg, I. Irfan,
M. Hammock, Y. B. Wang, C. Jia, X. Guo, Y. Gao, Z. Bao, Nano Lett.
2012, 12, 3574.
[18] T.-K. Hong, D. W. Lee, H. J. Choi, H. S. Shin, B.-S. Kim, ACS Nano
2010, 4, 3861.
[19] V. C. Tung, L.-M. Chen, M. J. Allen, J. K. Wassei, K. Nelson,
R. B. Kaner, Y. Yang, Nano Lett. 2009, 9, 1949.
[20] S. De, P. E. Lyons, S. Sorel, E. M. Doherty, P. J. King, W. J. Blau,
P. N. Nirmalraj, J. J. Boland, V. Scardaci, J. Joimel, ACS Nano 2009,
3, 714.
[21] Y. Miyata, K. Yanagi, Y. Maniwa, H. Kataura, J. Phys. Chem. C 2008,
112, 3591.
[22] P.-X. Hou, W.-S. Li, S.-Y. Zhao, G.-X. Li, C. Shi, C. Liu, H.-M. Cheng,
ACS Nano 2014, 8, 7156.
[23] H. W. C. Postma, M. de Jonge, Z. Yao, C. Dekker, Phys. Rev. B 2000,
62, R10653.
[24] A. Znidarsic, A. Kaskela, P. Laiho, M. Gaberscek, Y. Ohno,
A. G. Nasibulin, E. I. Kauppinen, A. Hassanien, J. Phys. Chem. C
2013, 117, 13324.
[25] D. Simien, J. A. Fagan, W. Luo, J. F. Douglas, K. Migler, J. Obrzut,
ACS Nano 2008, 2, 1879.
[26] B.-J. Lee, E.-C. Shin, G.-H. Jeong, Appl. Phys. A: Mater. Sci. Process.
2012, 107, 843.
[27] P.-X. Hou, B. Yu, Y. Su, C. Shi, L.-L. Zhang, C. Liu, S. Li, J.-H. Du,
H.-M. Cheng, J. Mater. Chem. A 2014, 2, 1159.
[28] K. Mustonen, P. Laiho, A. Kaskela, T. Susi, A. G. Nasibulin,
E. I. Kauppinen, Appl. Phys. Lett. 2015, 107, 143113.
[29] Y. Tian, M. Y. Timmermans, S. Kivistö, A. G. Nasibulin, Z. Zhu,
H. Jiang, O. G. Okhotnikov, E. I. Kauppinen, Nano Res. 2011, 4,
807.
[30] J. Huang, Q. Zhang, M. Zhao, F. Wei, Nano Res. 2009, 2,
872.
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
www.advancedsciencenews.com
[31] S. Corthals, J. Van Noyen, J. Geboers, T. Vosch, D. Liang, X. Ke,
J. Hofkens, G. Van Tendeloo, P. Jacobs, B. Sels, Carbon 2012, 50,
372.
[32] A. Magrez, J. W. Seo, V. L. Kuznetsov, L. Forró, Angew. Chem., Int.
Ed. 2007, 46, 441.
[33] Z. Li, Y. Xu, X. Ma, E. Dervishi, V. Saini, A. R. Biris, D. Lupu,
A. S. Biris, Chem. Commun. 2008, 3260.
[34] Q. Wen, W. Qian, F. Wei, Y. Liu, G. Ning, Q. Zhang, Chem. Mater.
2007, 19, 1226.
[35] Y. Tian, H. Jiang, J. van Pfaler, Z. Zhu, A. G. Nasibulin, T. Nikitin,
B. Aitchison, L. Khriachtchev, D. P. Brown, E. I. Kauppinen, J. Phys.
Chem. Lett. 2010, 1, 1143.
[36] M. S. Dresselhaus, G. Dresselhaus, R. Saito, A. Jorio, Phys. Rep.
2005, 409, 47.
[37] M. Dresselhaus, G. Dresselhaus, A. Jorio, J. Phys. Chem. C 2007,
111, 17887.
[38] A. Moisala, A. G. Nasibulin, D. P. Brown, H. Jiang, L. Khriachtchev,
E. I. Kauppinen, Chem. Eng. Sci. 2006, 61, 4393.
[39] K. Mustonen, P. Laiho, A. Kaskela, Z. Zhu, O. Reynaud,
N. Houbenov, Y. Tian, T. Susi, H. Jiang, A. G. Nasibulin, Appl. Phys.
Lett. 2015, 107, 013106.
Adv. Mater. Interfaces 2018, 5, 1801209 1801209  (10 of 10)
www.advmatinterfaces.de
[40] W. C. Hinds, Aerosol Technology: Properties, Behavior, and Measure-
ment of Airborne Particles, John Wiley & Sons, New York, NY 2012.
[41] K. K. Kim, J. J. Bae, H. K. Park, S. M. Kim, H.-Z. Geng, K. A. Park,
H.-J. Shin, S.-M. Yoon, A. Benayad, J.-Y. Choi, J. Am. Chem. Soc.
2008, 130, 12757.
[42] Y. Liao, H. Jiang, N. Wei, P. Laiho, Q. Zhang, S. A. Khan,
E. I. Kauppinen, J. Am. Chem. Soc. 2018, 140, 9797.
[43] A. Kaskela, P. Laiho, N. Fukaya, K. Mustonen, T. Susi, H. Jiang,
N. Houbenov, Y. Ohno, E. I. Kauppinen, Carbon 2016, 103,
228.
[44] J. L. Blackburn, T. M. Barnes, M. C. Beard, Y.-H. Kim, R. C. Tenent,
T. J. McDonald, B. To, T. J. Coutts, M. J. Heben, ACS Nano 2008, 2,
1266.
[45] D. S. Hecht, A. M. Heintz, R. Lee, L. Hu, B. Moore, C. Cucksey,
S. Risser, Nanotechnology 2011, 22, 169501.
[46] W. B. Liu, S. Pei, J. Du, B. Liu, L. Gao, Y. Su, C. Liu, H. M. Cheng,
Adv. Funct. Mater. 2011, 21, 2330.
[47] D. W. Shin, J. H. Lee, Y. H. Kim, S. M. Yu, S. Y. Park, J. B. Yoo, Nano-
technology 2009, 20, 475703.
[48] P. Laiho, K. Mustonen, Y. Ohno, S. Maruyama, E. I. Kauppinen,
ACS Appl. Mater. Interfaces 2017, 9, 20738.
© 2018 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim