Plasma-thermal purification and annealing of CNTs

1

Plasma-thermal purification and annealing of

carbon nanotubes

Asmus Meyer-Plath*, Guillermo Orts-Gil, Sergey Petrov, Franz Oleszak, Heinz-Eberhard

Maneck, Ilona Dörfel, Oskar Haase, Silke Richter, Reinhard Mach

BAM – Federal Institute for Materials Research and Testing, Division 6.5, Unter den Eichen

87, 12205 Berlin, Germany.

Abstract

We have developed a very fast and entirely gas-phase based purification technique for carbon

nanotubes (CNT) that allows removing metal and metal oxide impurities with high

effectiveness. CNT agglomerates from chemical vapor deposition (CVD) synthesis which

contained carbon encapsulated catalysts were injected into an atmospheric plasma torch. Very

high heating rates allow for quasi-instantaneous vaporization of catalyst particles. This way,

metal vapors are hyposized to break mechanically instable encapsulations and effuse from

incomplete ones faster than thermally induced graphitization stabilizes such particle

encapsulations. The ash content of multi-walled (MW) CNT samples was reduced to less than

15 % of the initial value within a few milliseconds. Also the metal content of single-walled

(SW) CNT agglomerates was significantly reduced. Repeated injection of CNT agglomerates

into the plasma torch resulted in higher-purity products of improved structural integrity and

increased oxidation resistance.

* Corresponding author. Fax +49 30 8104 1637. E-mail address: [email protected] (A. Meyer-Plath)

2

1. Introduction

During the last two decades, significant progress has been made on the synthesis of CNTs

with respect to growth efficiency and structural order. Impurities resulting from the CNT

synthesis are nonetheless still an important subject of contemporary research. They generally

deteriorate desired properties of CNTs and of derived products like composites. Much

research has been devoted to the removal of amorphous carbon, nanographites or fullerene-

like impurities by selective oxidation in liquid and gas phase [1]. Also plasma-based

approaches have been studied [2,3].

Although metal-free synthesis of CNTs is possible [4], metal catalysts increase the

conversion efficiency of hydrocarbons to nanotubular carbons and can give control over the

tube structure [5]. They are therefore key ingredients for the ongoing upscaling of CNT

production by fixed- and fluidized-bed CVD. During synthesis, catalysts react with carbon

precursors to tubular structures and encapsulating carbon overcoats. Together with metal

oxide supports they may attach to or become incorporated into the CNT product [6]. This is

problematic for many applications. Catalyst metals like nickel, cobalt or yttrium, are of

toxicological concern [7]. While un-encapsulated catalyst metals may be readily bio-available,

encapsulated ones may show metal ion leaching only on the long-term after (bio-)degradation

of their carbon shell. The removal of such encapsulated catalysts is a challenge and motivated

the development of the process presented here.

The lowest amount of metal impurities is currently obtainable by super-growth of vertically-

aligned CNTs on catalyst-coated substrates [8]. Contemporary mass production of multi-

walled CNTs by fluidized bed CVD synthesis generally results in 1-10 wt.% impurities [6].

Whereas the synthesis of single-walled CNTs by the arc or HiPCO method typically results in

a metal content of 20-30 wt.% [9]. In accord with the economic and scientific value of purified

CNTs, especially of SWCNTs, many different approaches have been developed to remove

3

metals and metal oxides [10–15].

The still most widely applied metal purification method is mineral acid washing, by which

incompletely encapsulated metal impurities can be readily converted to water soluble metal

halides. For carbon encapsulated impurities however, preceding oxidative treatment in

oxidizing acids or oxygen containing gas phases is necessary in order to perforate carbon

shells and render enclosed metals accessible to subsequent mineral acids dissolution [1,16].

Such oxidative treatment requires careful optimization, since it may cause CNT structure

defects and unintended chemical functionalization. In addition, byproducts like oxidized CNT

wall fragments can be formed. Such CNT-derived surfactants may be difficult to remove [17].

Since the worldwide production capacities for CNTs through CVD synthesis have

meanwhile surpassed the metric kiloton scale, acid purification concepts that were developed

on the lab scale have to prove their purification and cost-effectiveness for mass processing.

Especially acid purification and subsequent neutral washing steps for CNT agglomerates of

low apparent density, which are typical for industry-scale fluidized-bed mass production,

require handling large volumes of acids and waste waters as well as time-consuming filtering

and drying steps. Therefore entirely gas-phase-based metal purification concepts like the one

presented here, may be advantageous since they allow avoiding dry-wet-dry transfer steps.

The high thermal stability of CNTs under inert conditions allows thermal annealing in

graphite furnaces, sometimes supported by vacuum or halogen purification gases. It is well

known to be effective in metal removal, structure annealing and reduction of functional groups

[18]. The removal of nanosized CNT catalyst particles may be facilitated by melting point

depression [19]. Sufficiently high temperatures and long processing times from a few minutes

up to several hours may even remove incompletely encapsulated metals by diffusion processes

but are generally accompanied by – not always desired – graphitization of the CNTs and

carbonaceous impurities like catalyst coatings [20,21].

4

Here, a new approach to CNT purification by thermal annealing is presented that uses

thermal plasma processing. In contrast to furnace purification, the new approach uses

extremely high heating rates and allows processing times in the millisecond range.

2. Materials and methods

2.1 Materials

The performance of the new technique was studied for MWCNT and SWCNT agglomerates

from commercial fluidized-bed CVD synthesis with an ash content in the order of 2 wt. %.

MWCNT agglomerates of type “Baytubes® C150P” were purchased from Bayer

MaterialScience AG, Germany. The results of plasma-thermal purification were benchmarked

against a commercially metal-depleted MWCNT material of type “Baytubes® C150HP” from

the same company. The MWCNT agglomerates of 50-800 µm size exhibited a median

diameter of 250 µm and apparent density of about 160 kg/m³. The MWCNT were specified to

have a mean diameter of 13 nm and a BET surface of 250 m²/g. SWCNT agglomerates of type

“Elicarb® PR925” of comparable agglomerate size were purchased from Thomas Swan & Co.

Ltd., United Kingdom. The SWCNT were specified to have a mean diameter of 0.9-1.7 nm

and a BET surface of 700 m²/g.

2.2 Plasma processing

MWCNT and SWCNT agglomerates were dispersed at a controlled feeding rate of 100 g/h

into a process gas stream. The formed aerosol was injected into a thermal plasma torch

through a hollow probe, cf. Figure 1. The plasma torch was generated inductively by a 3-

winding coil around a 50 mm quartz tube which was supplied with 4 MHz radio frequency

power. Standard plasma torch operation parameters were 12 kW power at a gas flow of 78 slm

argon and 2.5 slm hydrogen. The equivalent thermodynamic energy of the plasma torch core

5

was estimated by calorimetric probe measurements to correspond to a gas temperature

approaching 10,000 K in the torch core. The hot process gas containing the processed CNTs

was cooled in a downstream water-cooled condenser. CNTs were filtered off the process gas

by a centrifugal separator and collected in an airtight container. In order to guarantee inert

conditions for the product and to avoid exposure of the laboratory staff to aerosols containing

nanoparticles, the processing setup was completely sealed by o-rings. Exhaust gases were

treated by a gas washer. At a fixed probe injection gas flow of 10 slm, the torch passage time

was varied by using injection probes of different bore diameter of 2, 5 and 10 mm. A gas flow

of 70.5 slm immersed the injection probe. The calculated particle exit speeds at the probe tip

were 13, 2 and 0.5 m/s, respectively called “fast”, “medium” and “slow” injection in the

following. The resulting particle speed inside the torch however was found to be affected by

thermal expansion of the process gas inside the hot plasma zone, see Section 3.2. The

residence time of agglomerates in the hot plasma zone visually appeared increased due to

more turbulent flow at lower injection speed.

2.3 Particle temperature and speed

An estimate of the average surface temperature of CNT agglomerates was obtained

spectrometrically. The optical emission of particles leaving the hot plasma zone was collected

by a biconvex lens of 50 mm focal length onto the end of an 8 µm quartz fiber that was

positioned below the visible emission of the plasma torch, cf. Figure 1. The fiber was

connected to a CCD spectrometer of type AvaSpec-2048-SPU, Avantes BV, The Netherlands.

The light of brightly glowing particles passing the line-of-sight was integrated for 200 ms. The

transmission function of the spectrometer was determined by collecting the emission spectrum

of a graphite rod that was heated inductively under argon. The surface temperature of the rod

was measured by a two-color ratio fiber optic pyrometer, assuming a graphite rod emission

6

coefficient of 90 %. The theoretical black body spectrum was fitted to the continuum radiation

part of the emission spectrum of the CNT agglomerates in the spectral range from 300-

950 nm.

Measurement of the lengths of trajectories of glowing particle on digital photographs taken

with an exposure time of 1/2000 s provided information of individual particle speeds.

2.4 Chemical and structural characterization

The residual mass was determined after ashing CNT samples in a furnace in air at 800 °C

for 4 h. Quantitative metal analysis of MWCNTs and SWCNTs was performed by electro-

thermal vaporization inductively coupled plasma optical emission spectroscopy (ETV-ICP

OES) and direct current (DC) arc OES analysis, respectively, using a TJA IRIS Advantage

DUO HR ICP OES spectrometer by Fischer Scientific GmbH, Germany. For ETV-ICP OES it

was coupled with an ETV 4000 by Spectral Systems, Germany. For DC arc OES, a Spectral

Systems DCA-301 direct current arc was used instead of the ICP burner [22]. The thermal

stability of the samples was determined by thermogravimetric analysis (TGA) in synthetic air

at a heating rate of 1 or 10 K/min. Raman spectra were acquired at 514.532 nm with a laser

power of 2 mW and laser spot size of 2 µm diameter on a LabRAM HR spectrometer by

Horiba Jobin Yvon GmbH, Germany. The measurements were taken at four different sample

spots. TEM investigations were performed on a JEOL JEM 2200-FS transmission electron

microscope operating at 200 kV equipped with an in-column omega-type energy filter. TEM

samples were prepared by immersion of Cu TEM grids coated with a holey amorphous carbon

film (S-147-3 Plano, Germany) into ethanolic suspensions of CNTs and solvent evaporation in

a dust-protected atmosphere.

7

3. Results and Discussion

3.1 CNT structure retention and oxidative stability

The extreme torch enthalpy of our device is in principle capable of evaporating metallic

impurities and destroying the CNT structure by graphite sublimation. CNT agglomerates were

found to leave the torch in a whitish glowing state. The spectroscopic observation of C2 Swan

band emissions below the plasma torch indicate that CNT agglomerates are treated close to the

carbon sublimation temperature. However, the TEM images in Figures 2 and 3 show that the

integrity of the tubular CNT structural can be preserved at sufficiently short torch passage

time. Depending on process conditions and number of processing cycles, significant structure

improvements are observable for MWCNTs using Raman spectroscopy in Figure 10. They

indicate an increase of the G band relative to the D band. Also the processed SWCNT samples

show small structural improvements according to the Raman spectra in Figure 11. The spectra

of processed SWCNT exhibit an additional peak at about 1850 cm-1. It has been attributed to

chain-like carbon material inside multi- or double-walled CNTs and is a characteristic of heat-

treated CNTs [23–25]. Jinno et al. reported disappearance of the peak after enhanced high

temperature annealing [23]. Here, the peak disappeared after repeated high temperature

annealing. According to Fantini et al., the resonance peak at 1850 cm-1 can result from the

coalescence of the inner tubes of neighboring double-walled CNT by linear carbon chains,

forming at special annealing temperatures [24]. This indicates the necessity of further work on

thermally induced diameter and (chiral) structure changes of CNTs. Also the differential TGA

curves in Figures 8 and 9 show strong improvements in oxidative stability of up to 125 K for

the MWCNT and 90 K for the SWCNT samples.

3.2 Purification effect

Optical emission spectra of CNT agglomerates were acquired for the three different

8

injection speeds. By adapting the black body temperature, a Planck emission spectrum was

fitted to the continuum radiation part of measured emission spectra. The reliability of this

approach is however restricted since the expected emission maximum at about 1050 nm was

not determined directly and spectral regions with C2-Swan band emissions from carbon

sublimation had to be omitted. This way CNT agglomerate temperatures were determined to

be approximately 2400, 2500 and 2700 °C for injection probes of 2, 5 and 10 mm inner bore,

corresponding to fast, medium and slow injection.

The average particle speeds inside the torch, as determined from track length on

photographs of glowing particles, were (21 ± 6), (13 ± 2) and (10 ± 1) m/s for injection probes

of 2, 5 and 10 mm inner bore. As discussed in Section 2.2, the speeds are deviating from the

calculated speeds of 13, 2 and 0.5 m/s at the tip of the injection probe. The transit times for the

hot plasma zone of about 150 mm length therefore lie in the order of 7, 12 and 15 ms,

respectively. Together with the average particle temperature estimates, particle heating rates

are expected to exceed 150,000 K/s. Such high heating rates could quasi instantaneously

vaporize volatile impurities and allow purifying metal-containing CNTs.

High-temperature annealing promotes both vaporization of metals as well as graphitization

of catalyst encapsulations. Figure 4 shows how quick the graphitization of carbon shells in a

SWCNT sample can proceed. Only 3 to 6 hot plasma zone passages of a few milliseconds

each were sufficient to form highly graphitic encapsulations. Therefore, if the temperature

during furnace annealing rises too slowly, any graphitization progress of encapsulations before

metal vaporization temperatures are reached will reduce the purification effectiveness.

A rapid vapor pressure build-up inside a closed encapsulation may break mechanically weak

carbon shells faster than a thermally induced reordering and graphitization may improve the

encapsulation’s integrity and stability. This way, optimized plasma-thermal annealing can

open a process window for the efficient removal of carbon-encapsulated catalysts.

9

For the present work, the purification effect was further optimized by separating metal

(oxide) vapors from the solid product by condensation in a downstream cold-wall reactor.

Comparison of plasma-thermally annealed sample residues after ashing in Figure 5 shows

that for slow MWCNT injection, the purification effect of a single torch passage approached

the purity level of commercially metal-depleted MWCNTs. After plasma processing of

MWCNTs, no signs of encapsulated catalysts were found during TEM analysis. Instead, open-

ended multi-walled tubes were observed, which were not observed in the untreated sample, cf.

Figure 2. This supports the hypothesis that carbon encapsulations may be broken by rapidly

vaporizing catalyst metals.

In order to assess the potential of the new approach after a possible future process

optimization and higher power input, MWCNT batches were processed repeatedly by re-

injecting processed material, each cycle called “torch passage” in the following. As ETV-ICP

OES analysis in Figure 6 shows, repeated plasma processing can further reduce the impurity

level. Most importantly, the content of the heavy metal cobalt can be reduced well below the

mg/g level.

Applying the process to SWCNTs from CVD synthesis at medium injection speed at 12 kW

plasma power resulted in a significantly lower purification effect than for MWCNTs, cf.

Figure 7. Only after several injection cycles the remaining metal content reached about 40 %

of the initial value. An increase in the plasma power to 20 kW showed that additional plasma

enthalpy can improve the purification efficiency. CVD synthesis of SWCNTs standardly

requires temperatures that are 500-700 K higher than for MWCNTs. The limited purification

effect for SWCNTs is therefore attributed to the stability of catalyst encapsulations in the

starting material. It depends on the carbon shell integrity, thickness and graphitization degree,

which generally increase with synthesis or post-processing temperature. Figure 4 shows

SWCNT catalyst particles that survived triple or sixfold plasma-thermal processing at 12 kW

10

power. They exhibit multilayer graphitic encapsulations that may be too stable to be broken by

internal pressure rise due to catalyst vaporization. This explains why for repeated processing

the purification effect per cycle was found to decrease in Figure 7.

4. Conclusion and Perspectives

Plasma-thermal processing of CNTs allows a very rapid removal of metal-containing

impurities. High processing rates are possible if sufficient plasma power density can be

provided. Although the boiling points of contained impurities like alumina, cobalt, iron and

magnesia are close or exceeding thermal stability predictions for CNTs [26], the product yield

was in the range of 80 to 95 %. This shows that CNT structure-preserving plasma purification

conditions were established. The thermal removal of metal impurities is assumed to be

facilitated by melting point depression of nanoscale particles [27]. In addition, plasma-

generated activated hydrogen species can chemically reduce metal oxides to elemental metals

of lower boiling point [28].

The structure of the carbon encapsulation of catalyst particles critically affects the

purification effectiveness. Mechanically weak encapsulations, which prevailed in the

MWCNT sample, were broken efficiently, whereas strong multilayer encapsulations, which

form during higher synthesis temperature of SWCNTs, can in principle withstand higher

internal pressures of vaporizing catalysts. As a consequence, encapsulated catalysts that stay

intact may even be further stabilized by (repeated) plasma-thermal processing due to

graphitization of encapsulations and enhancement of catalyst crystallinity cf. Figure 4. For

future progress, it will therefore be important to further increase the power density of the

plasma torch and to optimize the residence time. As many catalyst encapsulations as possible

should be cracked during the first plasma passage before graphitization promotes their

stability. This will require rapid catalyst vaporization at carefully optimized annealing

11

temperatures right below the CNT sublimation threshold. Since the experimental setup of the

present work was limited to 20 kW plasma power, additional process optimization should be

subject of future work.

Since plasma-thermal annealing exhibits various process parameters, there is plenty of room

for further research and improvements. The admixture of halogen-containing gases to further

improve metal volatilization and the addition or quenching with oxidizing process gases to

etch carbonaceous catalyst encapsulations and remove amorphous impurities are especially

promising [3]. Its high degree of effectiveness, flexibility and – being a continuous, entirely

gas phase-based process – in-line compatibility make plasma thermal annealing an interesting

alternative to conventional purification approaches.

Acknowledgements

The authors thank Sigrid Benemann, Angelika Dette, Jörg Friedrich, Sebastian Geier, Gundula

Hidde, Dietmar Neubert, Ludwig Plath and Sylvia Ziemann for their support. This work was

partially funded by the German Federal Ministry of Education and Research (BMBF), grant

number 03X0041D.

References

1. Hou P-X, Liu C, Cheng H-M. Purification of carbon nanotubes. Carbon. 2008

Dec;46(15):2003–25.

2. Morishita K, Takarada T. Scanning electron microscope observation of the purification

behaviour of carbon nanotubes. Journal of Materials Science. 1999 Mar 1;34(6):1169–

74.

12

3. Shahverdi A, Keun Su Kim, Alinejad Y, Soucy G. In situ removal of amorphous carbon

from single-walled carbon nanotubes synthesized by induction thermal plasma [Internet].

In: 2008 IEEE 35th International Conference on Plasma Science. Karlsruhe, Germany:

2008 [cited 2011 Aug 10]. p. 1.Available from:

http://ieeexplore.ieee.org/Xplore/login.jsp?url=http%3A%2F%2Fieeexplore.ieee.org%2F

iel5%2F4577664%2F4590586%2F04590886.pdf%3Farnumber%3D4590886&authDecis

ion=-203

4. Liu B, Ren W, Gao L, Li S, Pei S, Liu C, et al. Metal-Catalyst-Free Growth of Single-

Walled Carbon Nanotubes. Journal of the American Chemical Society. 2009 Feb

18;131(6):2082–3.

5. Yamada T, Namai T, Hata K, Futaba DN, Mizuno K, Fan J, et al. Size-selective growth

of double-walled carbon nanotube forests from engineered iron catalysts. Nat Nano. 2006

Nov;1(2):131–6.

6. MacKenzie KJ, Dunens OM, Harris AT. An Updated Review of Synthesis Parameters

and Growth Mechanisms for Carbon Nanotubes in Fluidized Beds. Industrial &

Engineering Chemistry Research. 2010 Jun 2;49(11):5323–38.

7. Jakubek LM, Marangoudakis S, Raingo J, Liu X, Lipscombe D, Hurt RH. The inhibition

of neuronal calcium ion channels by trace levels of yttrium released from carbon

nanotubes. Biomaterials. 2009 Oct;30(31):6351–7.

8. Hata K, Futaba DN, Mizuno K, Namai T, Yumura M, Iijima S. Water-Assisted Highly

Efficient Synthesis of Impurity-Free Single-Walled Carbon Nanotubes. Science. 2004

Nov 19;306(5700):1362 –1364.

9. Xu Y-Q, Peng H, Hauge RH, Smalley RE. Controlled Multistep Purification of Single-

13

Walled Carbon Nanotubes. Nano Letters. 2005 Jan 1;5(1):163–8.

10. Rinzler AG, Liu J, Dai H, Nikolaev P, Huffman CB, Rodríguez-Macías FJ, et al. Large-

scale purification of single-wall carbon nanotubes: process, product, and

characterization. Applied Physics A: Materials Science & Processing. 1998 Jul

15;67(1):29–37.

11. Thien-Nga L, Hernadi K, Ljubovic E, Garaj S, Forro L. Mechanical Purification of

Single-Walled Carbon Nanotube Bundles from Catalytic Particles. Nano Letters. 2002

Dec 1;2(12):1349–52.

12. Li X, Yuan G, Brown A, Westwood A, Brydson R, Rand B. The removal of encapsulated

catalyst particles from carbon nanotubes using molten salts. Carbon. 2006

Aug;44(9):1699–705.

13. MacKenzie K, Dunens O, Harris AT. A review of carbon nanotube purification by

microwave assisted acid digestion. Separation and Purification Technology. 2009 Apr

20;66(2):209–22.

14. Yumura M, Ohshima S, Uchida K, Tasaka Y, Kuriki Y, Ikazaki F, et al. Synthesis and

purification of multi-walled carbon nanotubes for field emitter applications. Diamond

and Related Materials. 1999 Mar;8(2-5):785–91.

15. Xu Y-Q, Peng H, Hauge RH, Smalley RE. Controlled Multistep Purification of Single-

Walled Carbon Nanotubes. Nano Letters. 2005 Jan 1;5(1):163–8.

16. Ebbesen TW, Ajayan PM, Hiura H, Tanigaki K. Purification of nanotubes. Nature. 1994

Feb 10;367(6463):519.

17. Hanelt S, Orts-Gil G, Friedrich JF, Meyer-Plath A. Differentiation and quantification of

14

surface acidities on MWCNTs by indirect potentiometric titration. Carbon. 2011

Aug;49(9):2978–88.

18. Huang W, Wang Y, Luo G, Wei F. 99.9% purity multi-walled carbon nanotubes by

vacuum high-temperature annealing. Carbon. 2003;41(13):2585–90.

19. Wang J, Duan H., Huang Z., Karihaloo B. A scaling law for properties of nano-

structured materials. Proceedings of the Royal Society A: Mathematical, Physical and

Engineering Science. 2006 May 8;462(2069):1355–63.

20. Chen J, Kuno A, Matsuo M, Tsukada T, Tamura T, Osato K, et al. Removal of entrapped

iron compounds from isothermally treated catalytic chemical vapor deposition derived

multi-walled carbon nanotubes. Carbon. 2008 Mar;46(3):391–6.

21. Andrews R, Jacques D, Qian D, Dickey EC. Purification and structural annealing of

multiwalled carbon nanotubes at graphitization temperatures. Carbon. 2001

Sep;39(11):1681–7.

22. Matschat R, Haßler J, Traub H, Dette A. Multielement trace determination in SiC

powders: assessment of interlaboratory comparisons aimed at the validation and

standardization of analytical procedures with direct solid sampling based on ETV ICP

OES and DC arc OES. Analytical and Bioanalytical Chemistry. 2005 Aug 4;383(7-

8):1060–74.

23. Jinno M, Ando Y, Bandow S, Fan J, Yudasaka M, Iijima S. Raman scattering study for

heat-treated carbon nanotubes: The origin of ≈1855 cm−1 Raman band. Chemical

Physics Letters. 2006 Jan 25;418(1–3):109–14.

24. Fantini C, Cruz E, Jorio A, Terrones M, Terrones H, Van Lier G, et al. Resonance

15

Raman study of linear carbon chains formed by the heat treatment of double-wall carbon

nanotubes. Phys. Rev. B. 2006 May 25;73(19):193408.

25. Ando Y, Zhao X. Synthesis of Carbon Nanotubes by Arc-Discharge Method. New

Diamond and Frontier Carbon Technology. 2006;16(3):123–37.

26. Begtrup G, Ray K, Kessler B, Yuzvinsky T, Garcia H, Zettl A. Probing Nanoscale Solids

at Thermal Extremes. Phys. Rev. Lett. [Internet]. 2007 Oct [cited 2011 Aug 15];99(15).

Available from: http://link.aps.org/doi/10.1103/PhysRevLett.99.155901

27. Buffat P, Borel J-P. Size effect on the melting temperature of gold particles. Phys. Rev.

A. 1976;13:2287.

28. Lyubochko VA, Malikov VV, Parfenov OG, Belousova NV. Reduction of aluminum

oxide in a nonequilibrium hydrogen plasma. J Eng Phys Thermophys. 2000

May;73(3):568–72.

16

Filter

Condensor 

Gas

CNT

Generator

OES

Figure 1: Schematic of the plasma-thermal torch setup with: (top to bottom) process gas inlets,

particle injection probe, RF inductor coil around the quartz tube, fiber probe for optical

emission spectroscopy (OES), quench gas inlets and cold-wall condenser unit.

17

Figure 2: TEM pictures showing Baytubes® C150P MWCNT samples: (A) an overview of

untreated C150P. The inset shows an embedded metal catalyst, STEM/EDX identified Co and

Mn; (B) C150P after plasma-thermal processing. Red arrows show MWCNTs with open ends;

(C) graphitic structure of C150P after plasma treatment; (D) open ended MWCNTs after

processing.

18

Figure 3: TEM pictures showing Elicarb® PR925 SWCNT samples before (A, B) and after (C,

D) plasma-thermal processing.

19

Figure 4: TEM pictures showing encapsulated catalysts in Elicarb® PR925 SWCNT

samples: (A) untreated; (B) after 3 and (C) after 6 plasma-thermal processing cycles.

20

C150P Fast Medium Slow C150HP Rem

ain

ing

ma

ssa

fte

ras

hin

g a

t 80

0°C

in

wt.

% 3.0%

2.5%

2.0%

1.5%

1.0%

0.5%

0.0%

Figure 5: Remaining ash mass of MWCNT agglomerates of type Baytubes® C150P after a

single plasma torch passage with varied injection speed compared to the starting material

C150P and to commercially metal-depleted C150HP after ashing at 800 °C in air. See text for

details and specification of fast, medium, and slow injection speed.

21

0

1

2

3

4

C150P 2006-11-15

(B2006)

1

(B3584)

2 Plasmabehandlungen

(B3584)

3

(B3586)

C150HP 2007-05-14

(B2022)

Co

nce

ntr

atio

nin

mg

/g (

wt.

‰)

Mg ‰

Al ‰

Co ‰

Mn ‰

C150P 1 passage 2 passages 3 passages C150HP MWCNT MWCNT

Figure 6: Remaining metal concentration in mg/g MWCNT mass after single, double and

triple plasma-thermal processing (torch passage) at medium injection speed (see text)

compared to the starting material Baytubes® C150P and commercially metal-depleted

C150HP as determined by ETV-ICP OES.

22

0

2

4

6

8

10

12

14

16

18

0 1 2 3 4 5 6

Fe concentration in mg/g (w

t.‰

)

Number of plasma torch passages

Low power ‐ Fe in mg/g

High power ‐ Fe in mg/g

Figure 7: Remaining iron concentration in mg/g SWCNT mass after different plasma-thermal

processing cycles (torch passage) at medium injection speed (see text) and low (12 kW, gray

line) as well as high (20 kW, black line) power conditions in comparison to the starting

material Elicarb® PR925 as determined by DC Arc OES.

23

-5

0

5

10

15

20

25

30

100 200 300 400 500 600 700 800 900

DT

G in

µg

/min

T in °C

DTG (B3657b)DTG (B3656b)DTG (B3655b)BT C150HPDTG (B2036)

12 kW

3 passages

2 passages

1 passage

C150HP MWCNT

C150P MWCNT

Figure 8: Differential thermogravimetry data of MWCNT samples after single, double and

triple plasma-thermal processing (torch passage) at 12 kW compared to the starting material

Baytubes® C150P and commercially metal-depleted C150HP acquired with a TGA heating

rate of 1 K/min in synthetic air.

24

-10

10

30

50

70

90

110

130

DT

G in

µg

/min

6 passages

3 passages

2 passages

1 passage

PR925 SWCNT

12 kW

-10

10

30

50

70

90

110

100 200 300 400 500 600 700 800 900

DT

G in

µg

/min

T in °C

3 passages

2 passages

1 passage

PR925 SWCNT

20 kW

Figure 9: Differential thermogravimetry data of SWCNT samples after plasma-thermal

processing (torch passage) at 12 kW (top) and 20 kW (bottom) plasma power compared to the

starting material Elicarb® PR925 acquired with a TGA heating rate of 10 K/min in synthetic

air.

25

0

0.001

0.002

0.003

0.004

0.005

0.006

0.007

1175 1275 1375 1475 1575 1675

Raman shift in 1/cm

<B2036> 0x <B3655b> 1x <B3656b> 2x <B3657b> 3x

GD

C150P 1 passage 2 passages 3 passages

No

rma

lize

dR

am

an

inte

ns

ity

Figure 10: Raman spectra of MWCNT samples after single, double and triple plasma-thermal

processing (torch passage) compared to the starting material Baytubes® C150P. The

measurements were taken at four different sample spots, averaged, scaled to an integral

intensity of 1 for the shown range and smoothed with a running mean of 2.5 wavenumbers bin

width.

26

0

0.005

0.01

0.015

0.02

0.025

1175 1275 1375 1475 1575 1675 1775 1875

Norm

alized Raman intensity

Raman Shift in 1/cm

PR925 SWCNT 1 passage 3 passages

Figure 11: Raman spectra of Elicarb® PR925 SWCNTs before and after single and triple

plasma-thermal processing (passage). The measurements were taken at four different sample

spots, averaged, scaled to an integral intensity of 1 in the range of 1175-1700 1/cm.