Langmuir 2008, 24, 2407-2411
2407
Growth of Diameter-Controlled Carbon Nanotubes from Fe-V-O Nanoparticles Size-Classified by Ligand-Exchanged Fractional Precipitation Itaru Gunjishima,* Takashi Inoue, and Atsuto Okamoto Toyota Central R&D Laboratories, Inc., Nagakute, Aichi, 480-1192, Japan ReceiVed October 3, 2007. In Final Form: December 12, 2007 Colloidal Fe-V-O nanoparticles prepared as carbon nanotube (CNT) growth catalysts were precisely size-classified by fractional precipitation. Furthermore, the classification ability was improved by the fractional precipitation after ligand exchange process, which allowed us to obtain narrower size distributions of nanoparticles. CNTs were grown from the nanoparticles in order to investigate the dependence of diameter distribution of CNTs on that of nanoparticles. The results show that the diameter distribution of CNTs grown from classified nanoparticles was narrower than that of CNTs grown from as-prepared nanoparticles.
1. Introduction Nanoscale colloidal particles have been attracting a great deal of recent attention because they exhibit unique properties, differing substantially from those of the corresponding bulk solids. For example, single magnetic domain structures in ferromagnetic nanoparticles are useful for high-density magnetic data storage applications.1,2 The fluorescence emission phenomenon arising from the quantum size effect3 can be used for display phosphor applications, and the superlattice structure of the nanoparticles which show surface plasmon resonance absorption in the optical wavelength region can be used for optical device applications.4,5 In addition, transition metal nanoparticles have recently been used as catalysts for the growth of carbon nanotubes (CNTs),6-12 which are expected to be used as wiring materials for future large-scale integration,6 fabrics,13 and radiation materials14 owing to their excellent electrical, mechanical, and thermal properties.15 In almost all of these applications of nanoparticles, precise control of the diameter and narrow diameter distribution of the nanoparticles are required, since the characteristic properties at the nanoscale are strongly dependent on the size of the nanoparticles.3 As for the CNT catalyst, it is expected that the * TEL: +81-561-71-7092,
[email protected].
FAX:
+81-561-63-6156,
E-mail:
(1) Hyeon, T.; Chung, Y.; Park, J.; Lee, S. S.; Kim, Y. W.; Park, B. H. J. Phys. Chem. B 2002, 106, 6831-6833. (2) Leslie-Pelecky, D. L.; Rieke, R. D. Chem. Mater. 1996, 8, 1770-1783. (3) Murray, C. B.; Norris, D. J.; Bawendi, M. G. J. Am. Chem. Soc. 1993, 115, 8706-8715. (4) Shimizu, T.; Teranishi, T.; Hasegawa, S.; Miyake, M. J. Phys. Chem. B 2003, 107, 2719-2724. (5) Teranishi, T.; Hasegawa, S.; Shimizu, T.; Miyake, M. AdV. Mater. 2001, 13, 1699-1701. (6) Sato, S.; Kawabata, A.; Kondo, D.; Nihei, M.; Awano, Y. Chem. Phys. Lett. 2005, 402, 149-154. (7) Li, Y.; Kim, W.; Zhang, Y.; Rolandi, M.; Wang, D.; Dai, H. J. Phys. Chem. B 2001, 105, 11424-11431. (8) Cheung, C. L.; Kurtz, A.; Park, H.; Lieber, C. M. J. Phys. Chem. B 2002, 106, 2429-2433. (9) Fu, Q.; Huang, S.; Liu, J. J. Phys. Chem. B 2004, 108, 6124-6129. (10) Choi, H. C.; Kim, W.; Wang, D.; Dai, H. J. Phys. Chem. B 2002, 106, 12361-12365. (11) Gunjishima, I.; Inoue, T.; Yamamuro, S.; Sumiyama, K.; Okamoto, A. Carbon 2007, 45, 1193-1199. (12) Gunjishima, I.; Inoue, T.; Yamamuro, S.; Sumiyama, K.; Okamoto, A. Jpn. J. Appl. Phys. 2007, 46, 3700-3703. (13) Zhang, M.; Fang, S.; Zakhidov, A. A.; Lee, S. B.; Alieve, A. E.; Williams, C, D.; Atkinson, K. R.; Baughman, R. H. Science 2005, 309, 1215-1219. (14) Huang, H.; Liu, C.; Wu, Y.; Fan, S. AdV. Mater. 2005, 17, 1652-1656. (15) Ajayan, P. M. Chem. ReV. 1999, 99, 1787-1800.
diameter distribution of CNTs is eventually reflected by the diameter distribution of the nanoparticles.8,9 Therefore, it is important to obtain nanoparticles with the desired diameters with small standard deviations. However, the various synthesis processes have their limitations for synthesizing nanoparticles of uniform size by only optimizing the synthesis conditions. Therefore, size classification of the nanoparticles after the synthesis process is important. Here, we report a method for precise size classification of the nanoparticles to obtain much narrower size distributions using a centrifugal fractional precipitation, along with controlling the thickness of the surfactants which cap the nanoparticles. We then grew CNTs of narrower diameter distribution using these nanoparticles as catalysts. 2. Experimental Methods 2.1. Nanoparticle Synthesis. The synthesis method for the FeV-O nanoparticles is similar to that described in ref 11. Fe(acetylacetonate)3 (0.9 to 0.75 mmol), VO(acetylacetonate)2 (0.1 to 0.25 mmol), 1,2-hexadecanediol (7 mmol), oleic acid (3 mmol), oleylamine (3 mmol), and dioctyl ether (20 mL) were stirred at room temperature for 1 h under an Ar gas flow, slowly heated, and kept at 130 °C for 30 min. The mixture was then rapidly heated and kept at 250 °C for 30 min under reflux. After the reaction solution was cooled to room temperature, a sufficient amount of ethanol was added in order to precipitate the nanoparticles. The nanoparticles were separated from solvent by centrifugation (13 000 g for 5 min) and were then redispersed in hexane. 2.2. Ligand Exchange of Nanoparticles. Ligand exchange of nanoparticles from C18 (carbon chain length ) 18) to C6 (carbon chain length ) 6) was performed for some portions of the nanoparticles as follows. Caproic acid (20 mmol) and hexylamine (20 mmol) were mixed with the nanoparticle-dispersed hexane, and the mixture was heated under reflux for 1 h at 60 °C. After the mixture was cooled to room temperature, the nanoparticles were rinsed and redispersed in hexane in a similar manner to that described above. 2.3. Size Classification and Evaluation of Nanoparticles. The as-synthesized nanoparticles and the ligand-exchanged nanoparticles were size-classified by a centrifugal fractional precipitation method. In typical fractional precipitation, a hexane solution of the particles was mixed with the ethanol until the solution started to cloud. Then, the mixed solution was centrifuged (13 000 g for 5 min), resulting in a small amount of nanoparticle precipitation. After decanting the supernatant liquid, the precipitated nanoparticles were redispersed in hexane. Subsequently, a small amount of ethanol (1 mL) was
10.1021/la703044t CCC: $40.75 © 2008 American Chemical Society Published on Web 02/09/2008
2408 Langmuir, Vol. 24, No. 6, 2008
Gunjishima et al.
added to the decanted supernatant liquid, and the mixture was centrifuged again in order to obtain further precipitation. Keeping the supernatant each time, the centrifugation and ethanol addition processes were repeated several times. Then, the nanoparticles were observed by transmission electron microscopy (TEM). For preparation of samples for TEM, drops of the nanoparticle suspension were placed on TEM grids which have thin carbon support film, and the grids were dried under vacuum. The average diameters of the as-synthesized and fractionated nanoparticles and standard deviations were estimated by measuring about 80 nanoparicles of each TEM image. The measuring direction was integrated in a TEM image for all nanoparticles including nonspherical shape nanoparticles, and the diameter of each nanoparticle was determined from maximum measurement length in this measuring direction. 2.4. Synthesis and Characterization of Carbon Nanotubes Grown from Nanoparticles. Cubic mesoporous silica films were used as substrates for the growth of CNTs in the present study. The substrate was dipped in the nanoparticle-dispersed hexane solution, and a monolayer of the nanoparticles was formed on a substrate as in our previous study11 by adjusting the dipping solution concentration. The surfactants coated on the nanoparticles were then removed by a 10 min plasma pretreatment to improve the catalytic activity of the nanoparticles. CNTs were grown using thermal CVD. The substrates were heated to 650 °C and kept at this temperature for 10 min. C2H2 (5 sccm) and H2 (90 sccm) flow was maintained during CNT growth, and the pressure was maintained at 266 Pa. The diameters of the CNTs were measured by TEM.
3. Results and Discussion 3.1. Size Classification of Nanoparticles by Centrifugal Fractional Precipitation Method. Figure 1a1 shows a typical TEM image of as-synthesized Fe-V-O nanoparticles on a TEM grid. In this figure, the spaces between the nanoparticles indicate the presence of surfactants which cap and stabilize the nanoparticles. The diameter histograms of the as-synthesized nanoparticles are shown in Figure 1a2. The average diameter (Ave.) of the nanoparticles and its standard deviation (σ) were estimated to be 4.0 and 0.87 nm, respectively. The standard deviation of about 0.7 to 1.0 nm is the typical value that we obtain by our liquid-phase synthesis process. To obtain nanoparticles with narrow size distributions, separation of the nucleation and growth periods during synthesis is important.16 However, complete separation of these periods in the actual synthesis process is quite difficult, and this difficulty becomes a significant problem in fabricating small nanoparticles with narrow diameter distributions. This is because the nanoparticles which nucleate in the initial stage start growing first compared to those that nucleate later, and the diameter difference due to this factor cannot be reduced even if the average diameter is decreased by shortening the growth period.16 This results in a large percentage of standard deviation to average diameter at small diameters. Figures 1b1, c1, d1, and e1 show TEM images of fractionated nanoparticles at each fraction number, and Figures 1b2, c2, d2, and e2 show their corresponding respective diameter histograms. The assynthesized nanoparticles with a diameter of 4.0 ( 0.87 nm were size-classified into fractions of nanoparticles with diameters of 4.9 ( 0.52 nm, 4.6 ( 0.57 nm, 4.7 ( 0.27 nm, 4.3 ( 0.36 nm, 4.0 ( 0.35 nm, and 3.6 ( 0.38 nm, which shows that the average diameter of the fractionated nanoparticles became narrower by repeating the centrifugal precipitation with stepwise addition of ethanol. The larger nanoparticles were the first to precipitate out, since the nanoparticle settling velocity is proportional to the square of (16) Sugimoto, T.; Dirige, G.; E.; Muramatsu, A. J. Colloid Interface Sci. 1996, 182, 444-456.
Figure 1. TEM images and diameter histograms of nanoparticles (a) before and (b-e) after classification.
the nanoparticle diameter; the particle settling velocity is given by V ) Dp2(Fp - Ff)g/18η (Stokes’ law),17 where g is the acceleration of gravity, Dp is the diameter of the particle, Fp is the density of the particle, Ff is the density of the fluid, and η is the viscosity of the fluid. The reason for the nanoparticles precipitating by the addition of ethanol to hexane may be a change in the polar character of the solvent that determines the dispersibility of nanoparticles. The TEM images of the fractionated particles showed self-organization of nanoparticles, indicating uniformity of the nanoparticles. The standard deviations of diameter of the fractionated nanoparticles were about half the magnitude of that of the as-synthesized nanoparticles. As mentioned above, the nanoparticles used in this study are capped by surfactants. Therefore, the thickness of the surfactant layer must be added to the nanoparticle diameter to estimate the nanoparticle settling velocity by Stokes’ law. For example, when the core size decreases from 3 to 2 nm (-33%), the change in the total nanoparticle size is 6.2 () 3 + 1.6 + 1.6) nm to 5.2 () 2 + 1.6 + 1.6 ) nm (-16%) taking into account the surfactant thickness (ca. 1.6 nm () 3.2 nm/2), 3.2 nm is measured value (17) Tseng, W. J.; Li, Mater, S. Y. J. Process Technol. 2003, 142, 408-414.
Growth of Diameter-Controlled CNTs
Langmuir, Vol. 24, No. 6, 2008 2409
Figure 2. TEM images of nanoparticles and correspoding histograms of spacings between the nanopraticles (a) before and (b) after C18 f C6 surfactant exchange.
of spacing between the nanoparticles in Figure 1). Thus, the presence of surfactants may reduce the settling velocity difference between a large particle and a small particle, which may cause degradation in the classification ability. From Stokes’ law, we can expect that, by decreasing the thickness of the surfactant layer, the settling velocity difference between the large particle and the small particle becomes large (the classification ability will be improved). We attempted to improve the ability for classification by exchanging surfactants from C18 to the shorter C6. Figure 2a1 and b1 are TEM images of the nanoparticles before and after the surfactant exchange, respectively, and Figure 2a2 and b2 show their corresponding histograms of spacings between the nanoparticles. The decrease in the spaces between the nanoparticles in Figure 2b suggests that the surfactants were successfully exchanged to the shorter one. Since the influence of the surfactant thickness on the classification ability is thought to be significant as the size of the nanoparticle becomes smaller, we compared the difference in the classification precision for relatively smaller nanoparticles as shown in Figure 3a. Figure 3b1 and c1 shows TEM images of fractionated nanoparticles capped by the C18 and C6 surfactants, respectively. Figure 3b2 and c2 shows the corresponding diameter histograms. Fraction numbers which have almost the same average diameter were selected for comparison (fractions no. 5 for C18 and no. 8 for C6). From comparison of the histograms of nanoparticles capped by the C18 and C6 surfactants, it is clear that the standard deviation of the diameter of fractionated nanoparticles capped by the C6
surfactant (2.6 ( 0.30 nm) is about 30% narrower than that of the nanoparticles capped by the C18 surfactant (2.8 ( 0.44 nm), suggesting that the classification ability was successfully improved by decreasing the surfactant thickness. Although a clear difference in the space between the nanoparticles was not observed due to the difficulty in self-organization for the small nanoparticles, the surfactants are thought to be exchanged from C18 to C6 since the same ligand-exchange treatment was performed. In contrast, shorter surfactants compared to C6, such as C4 and C2, were insoluble in nanoparticle-dispersed hexane solution, and we could not exchange to them. It seems that a certain length of alkyl chain is needed to dissolve in hexane solution. 3.2. CNT Growth from the Nanoparticles before and after Fractional Precipitation. Figure 4a,b shows representative TEM images of CNTs grown from as-synthesized nanoparticles and size-classified nanoparticles after ligand-exchange process, respectively. Both samples mainly contain multiwalled CNTs; however, the number of layers seems to be relatively large in the CNT sample grown from as-synthesized nanoparticles. Figure 5a,b shows diameter histograms of nanoparticles before and after classification (with C6 surfactants) and CNTs grown from these nanoparticles. These histograms were obtained by measuring about 100 objects in several images taken from different positions. CNTs were grown using nanoparticles with almost the same average diameter and different magnitude of standard deviation to confirm the effect of sharpening size distribution. As shown in the figures, although average diameter and standard deviation
2410 Langmuir, Vol. 24, No. 6, 2008
Gunjishima et al.
Figure 3. TEM images and diameter histograms of nanoparticles (a) before and after classification with (b) C18 surfactant and (c) C6 surfactant. The fraction numbers of sample (b) and (c) are no. 5 and no. 8, respectively.
Figure 4. TEM images of CNTs grown from (a) as-prepared and (b) size-classified nanoparticles.
of each CNTs are larger than those of original nanoparticles, magnitude correlation of the standard deviation of CNTs corresponds with that of the standard deviation of original nanoparticles (before classification, 3.5 ( 0.63 nm (nanoparticles) f 4.7 ( 1.2 nm (CNTs); after classification, 3.6 ( 0.31 nm (nanoparticles) f 4.1 ( 0.67 nm (CNTs)). Aggregation of nanoparticles might not have been completely suppressed during the CNT growth, and this has caused the larger average diameters and standard deviation of CNTs. To obtain CNTs whose diameter distribution is closer to that of the original nanoparticles, further optimization of growth conditions will be necessary. However, it has become clear that the sharpening process of the diameter
Figure 5. Diameter histograms of nanoparticles (NPs) (a) before and (b) after classification with C6 surfactant, and CNTs grown from these nanoparticles.
Growth of Diameter-Controlled CNTs
distribution of catalytic nanoparticles shown in this study can be an effective method to obtain CNTs of uniform diameter.
4. Conclusions As-prepared Fe-V-O nanoparticles were precisely sizeclassified into fractions of narrow distributions by centrifugal fractional precipitation along with control of the solvent composition. Nanoparticles covered by a shorter surfactant than the as-synthesized nanoparticles were prepared by a ligand exchange process and subsequently were classified in the same way. Diameter distribution evaluations using TEM showed that the classification ability was successfully improved by decreasing the surfactant thickness. The standard deviation of the diameter
Langmuir, Vol. 24, No. 6, 2008 2411
of CNTs grown from classified nanoparticles which have narrow diameter distribution showed a smaller value than that of CNTs grown from as-synthesized nanoparticles. We believe that this nanoparticle classification method is also useful for other nanoparticle applications. Acknowledgment. The authors would like to thank our collaborator Ms. K. Watanabe at Toyota CRDL for assisting us with all the experiments, and Dr. S. Yamamuro and Prof. K. Sumiyama at the Nagoya Institute of Technology for instructing us on the synthesis process of nanoparticles. LA703044T
