- Email: [email protected]
Kish graphite as a magnetic field sensor
3266
Letters to the Editor / Carbon 42 (2004) 3251–3272
Kish graphite as a magnetic field sensor Yutaka Kaburagi *, Yuichiro Asano, Yoshihiro Hishiyama Faculty of Engineering, Musashi Institute of Technology, 1-28-1, Tamazutsumi, Setagaya-ku, Tokyo 158-8557, Japan Received 14 May 2004; accepted 6 August 2004
Keywords: A. Carbon films, Highly oriented graphite; D. Defects, Galvanomagnetic properties
The materials showing notable magnetoresistance effect (MRE) and Hall effect have been applied to magnetic field sensors. As materials for the static field sensor due to MRE, semiconductor compounds, such as InSb, InAs and GaAs, and ferromagnetic metals, such as Ni–Co and Ni–Fe, have been used. The semiconductor sensors exhibit positive MRE and can measure magnetic fields between 106 and 102 T. The ferromagnetic sensors exhibit negative MRE and can only measure low magnetic fields down to 1010 from 102 T. On the other hand, highly oriented graphite materials, such as high quality natural graphite (NG), Kish graphite (KG) and highly oriented pyrolytic graphite (HOPG) are known to exhibit large positive magnetoresistance [1–4]. However, such graphite materials have not been used as MRE sensors. Transverse magnetoresistance Dq/q0 is expressed as D q/q0 = [q(B) q(0)]/q(0), where q(0) and q(B) are the resistivity of the sensor at zero-magnetic field and that at a magnetic field B applied perpendicular to the current flowing in the sensor, respectively. Graphite is a semimetal and has almost equal densities of electrons and holes. For such material, Dq/q0 is approximately proportional to the square of mean mobility pffiffiffiffiffiffiffiffiffi lð¼ le lh Þ, i.e., Dq/q0 ffi l2B2 = lelhB2, where le is the mean electron mobility and lh is the mean hole mobility [1,3–5]. le and lh are very sensitive to the presence of lattice defects and impurities, and decrease with increasing their densities in the sensor. Therefore, the magnitude of Dq/q0 depends on the crystallinity of graphite sample, and high quality graphite with very low defect density shows very large Dq/q0. KG is a graphite flake precipitated from molten iron at high temperatures [4,6]. When the KG flakes are produced at a very high temperature as high as that at
*
Corresponding author. Tel./fax: +81 3 5707 2204. E-mail addresses: [email protected] (Y. Kaburagi).
0008-6223/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2004.08.006
which iron evaporates and purified carefully, they have much higher crystallinity than that of HOPG and exhibit a nature similar to single crystal graphite [2,6–8]. A high quality KG sample, therefore, exhibits much larger Dq/q0 than that of HOPG, and comparable to that of a high quality NG [1,4,8,9]. In the present study, a static magnetic field sensor was tried to prepare using the KG flakes showing Dq/q0 higher than 100% at a field of 1 T at room temperature. Dq/q0 can be denoted as Dq/q0 = DV/V, where V is the voltage of the sensor at zero-field and DV is the voltage change induced by magnetic field. Therefore, the magnitude of B can be represented as that of DV with a constant electrical current supplied to the sensor. For a rod-shaped sensor of rectangular cross-section with the width w and thickness t, DV is given by DV = [q(B) q(0)](I‘/wt), where ‘ is the distance between the two electrodes for the voltage measurement and I is the electric current. DV can be obtained as a larger value with a larger ‘ and smaller w and t at a constant I. In the case of KG samples, ‘ is limited because they are small flakes and I should not be large for suppressing Joule heating which rises to fluctuation of DV. Therefore, reducing w and t is important factor to increase DV, but a careful treatment is necessary to reduce w and t, i.e. without increase of the defect density of the sample. The sizes of the KG samples prepared in the present experiment were those with 2 mm width and lower than 0.3 mm in thickness. The original KG flakes have irregular shapes, with their approximate sizes of 15 mm · 8 mm · 0.5 mm. Each of the flakes was sandwiched between artificial graphite plates and glued to them. The flakes sandwiched by artificial graphite plates were cut into segments in rectangular shape with 6 mm · 2 mm in size, and then KG specimens were separated by dipping the segments into the solvent for the glue. Thin samples with flat surfaces and various values in thickness were prepared by peeling off carefully the both surfaces of the rectangular KG specimens using adhesive tape. The KG samples thus
Letters to the Editor / Carbon 42 (2004) 3251–3272
3267
obtained were 0.256, 0.146, 0.080, 0.050 and 0.040 mm in thickness and are denoted hereafter as KG followed by the thickness, such as KG-0.256. For each sample, V was measured at zero-magnetic field, and then DV was measured at a magnetic field after V had been offset electrically. The field dependence of DV (i.e., D V/V = Dq/q0 in a constant V) was measured at fields up to 1.0 T in the field perpendicular to the sample surface at room temperature (about 20 C). For the measurements, four-terminal-method was performed and a constant current of 50 mA flowed. The DV measurement was also made in the reverse direction of the field to eliminate Hall voltage. However, the Hall voltage appeared is very small and negligible in comparison
with the voltage due to MRE. In addition, to obtain larger DV in the low field region, a commercially available neodymium disk magnet with 4 mm / and 2 mm thick was used as a bias field magnet because D q/q0 is proportional to B2 in the low magnetic field region as mentioned above. Namely, each KG sensor was equipped with the bias magnet after above measurements. The bias field applied to each sensor was 130 mT. It should be noted that the bias field direction has to coincide with the field direction to be measured. We obtain wrong value of DV if the bias field is opposite in direction to the measuring field. In Fig. 1, the field dependence of Dq/q0 at fields up to 1 T at room temperature (a) and the low field behavior
Fig. 1. Field dependence of transverse magnetoresistance at room temperature at fields up to 1.0 T (a) and 0.2 T (b) for KG samples with various values in thickness and without bias magnet.
Fig. 2. Field dependence of DV at fields up to 1 T (a) and 0.2 T (b) at room temperature for KG samples with various values in thickness without bias magnet.
3268
Letters to the Editor / Carbon 42 (2004) 3251–3272
(b) are shown for the KG samples with various values in thickness. These KG samples can measure much higher fields than 1 T according to Dq/q0 ffi l2B2. KG-0.256 is a relatively thick sample obtained by just removing irregular surfaces on both sides of one of the rectangular KG specimens and exhibits a positive Dq/q0 larger than 100% at a field of 1 T, indicating highly crystallized nature. Dq/q0 changes randomly with sample thickness as shown in Fig. 1. Defects may be introduced by peeling off the sample surfaces. The defect density or crystallinity of sample does not depend on the sample thickness, rather it depends on the peeling off condition, crystallinity of the original KG specimen and also crystallinity of the remained part after peeling off because KG flakes are polycrystals and aggregates of oriented large crystal grains [10]. Thin samples of KG-0.050 and KG-0.040 exhibit larger Dq/q0 than others, which means that defects are not introduced remarkably by reducing their thickness and the parts with good crystal condition are remained. For all samples, the fluctuation of Dq/q0 was slight on the temperature change of several degrees around 20 C, and the reproducibility of Dq/q0 measured at a constant magnetic field on different days was good, less than 2%. Fig. 2 shows the field dependence of DV at fields up to 1 T at room temperature (a) and the low field behavior (b) for the KG samples with various values in thickness. At a constant magnetic field, DV increases with decreasing sample thickness, and is 18 lV at a field of 0.1 T for KG-0.040. If the defect density increases with reducing the sample thickness t, Dq/q0 decreases, i.e., q(B) q(0) decreases. However, DV increases when the rate of decrease in t surpasses that in q(B) q(0) be-
Fig. 4. DV versus magnetic field plots at room temperature for KG samples equipped with bias magnet.
cause of DV = [q(B) q(0)](I‘/wt) as written above. For each sample, D V, as well as D q/q0 shown in Fig. 1, is roughly proportional to B2. Fig. 3 shows the field dependence of DV at fields up to 0.2 T at room temperature for the KG samples equipped with the bias magnet. At a constant magnetic field, DV increases remarkably with increasing B because of the bias magnet equipped to the samples, and each curve is closer to a straight line than that without bias magnet. In Fig. 4, the DV versus B plots are shown. KG-0.040 equipped with the bias magnet, being the most sensitive field sensor obtained in this study, exhibits a sensitivity of about 0.8 lV/mT around 0.1 T with a constant current of 50 mA at room temperature. The sensitivity enables us to measure magnetic field down to the order of mT using a conventional volt-meter. There is a possibility to measure magnetic field in the order of 104 T by reducing further the thickness and width of the KG sample, though the sensitivity will still be inferior to the semiconductor and ferromagnetic sensors.
References
Fig. 3. Field dependence of DV at fields up to 0.2 T at room temperature for KG samples equipped with bias magnet.
[1] Soule DE. Magnetic field dependence of the Hall effect and magnetoresistance in graphite single crystals. Phys Rev 1958;112(3):698–707. [2] Hishiyama Y, Ono A, Tsuzuku T, Takezawa T. Galvanomagnetic properties of well-oriented graphite in relation to the structural imperfections. Japn J Appl Phys 1972;11(7):958–63. [3] Kreps LW, Woollam JA. Temperature and field dependencies of galvanomagnetic effects in graphite. Carbon 1977;15: 403–09. [4] Kaburagi Y. Magnetoresistance of highly oriented graphite at 77 K and 4.2 K. J Phys C: Solid State Phys 1982;15:5425–40.
Letters to the Editor / Carbon 42 (2004) 3251–3272 [5] Dresselhaus MS, Dresselhaus G, Sugihara K, Spain IL, Goldberg HA. Graphite fibers and filaments. Berlin: Springer-Verlag; 1988. p. 202–216. [6] Inagaki M. New Carbons—Control of structure and functions. Amsterdam: Elsevier; 2000. p. 30–39. [7] Yoshida A, Hishiyama Y. Electron channeling effect on highly oriented graphites—Size evaluation and oriented mapping of crystals. J Mater Res 1992;7(6):1400–5. [8] Kaburagi Y, Yoshida A, Hishiyama Y. Microtexture of highly crystallized graphite as studied by galvanomagnetic properties
3269
and electron channeling contrast effect. J Mater Res 1996;11(3): 769–78. [9] Hishiyama Y, Kaburagi Y. Relationship between residual resistivity ratio qRT/q4.2 K and transverse magnetoresistance in highly oriented graphite. Tanso 1988;13:394–9. in Japanese. [10] Kaburagi Y, Yoshida A, Hishiyama Y. Microtexture of highly crystallized graphite as studied by galvanomagnetic properties and electron channeling contrast effect. J Mater Res 1996;11(3): 769–78.
Filling of multi-walled carbon nanotubes with tin(IV) oxide Liping Zhao, Lian Gao
*
The State Key Laboratory of High Performance Ceramics and Superfine Microstructure, Shanghai, Institute of Ceramics, Chinese Academy of Sciences, 1295 Ding Xi Road, Shanghai 200050, PR China Received 1 March 2004; accepted 11 August 2004
Keywords: A. Carbon nanotubes; B. Chemical treatment
Multi-walled carbon nanotubes (MWNTs) have attracted much interest since their discovery owing to their high mechanical strength and electrical properties [1,2]. Another fascinating aspect of carbon nanotubes is their cavities. Only a few nanometers in diameter and a few micrometers in length, such a cavity should allow for interesting nanoscale experiments, if it can be filled with other materials in a systematic way. Theoretical studies suggest that the introduction of foreign material into the inner cavities of MWNTs may enhance or modify the properties of the resulting composite materials [3,4]. The procedure to fill carbon nanotubes can be classified in two groups: (a) the chemical method [5–7], and (b) the physical method [8–10]. In the physical method, no solvent is used in the process and opened tubes are directly immersed in the molten material whereby the liquid is driven into the tube by capillary forces [7,10,11]. Thus, the physical method is more restrictive in the material choice and the ability of such materials to be so incorporated depends on three characteristics [12]: (i) they must ÔwetÕ the capillaries and have surface tensions below a threshold value in the range 100– 200 mN m1 [13]; (ii) their overall melting temperature must be low enough to preclude thermal damage to the MWNT; and (iii) they must not attack MWNTs *
Corresponding author. Tel.: +86 24 2397 1863; fax: +86 24 2389 1320 E-mail address: [email protected] (L. Gao). 0008-6223/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2004.08.009
chemically. The wet chemical method consists of the opening of nanotube ends and the inclusion of salt solution into the opened tubes. The main advantage of the wet chemical approach is its flexibility and the level of experimental control, so that a wider variety of materials can be introduced into carbon nanotubes than the physical method. SnO2 is one of the most important and useful semiconductor materials with a large band gap, which has been extensively used for gas sensors [14], transparent conducting electrodes [15], transistors, and solar cells [16]. Sloan et al. [17] reported a two-step wet chemical method using SnCl2 as a precursor for the filling of MWNTs with SnO nanoparticles in high yield, involving (a) opening and (b) filling. Unfortunately, they still need to dissolve SnCl2 in hot concentrated HCl solution first and the solid-state SnO is thermodynamically unstable with respect to b-Sn and SnO2. The decomposition reaction of 2SnO ! SnO2 + Sn was reported to occur starting at a temperature as low as 370 C [18]. Here we use a weak acidic SnCl2 solution as precursor for the filling of MWNTs with SnO2 nanocrystals. The addition of a small amount of acid is only for reducing the hydrolysis of SnCl2. MWNTs prepared by the catalytic decomposition of CH4 were kindly provided by Shenzhen Nanotech Port Ltd. Co. (Shengzhen, China) and used as received. The inner and outer diameters of the MWNTs measured by transmission electron microscopy (TEM,
Letters to the Editor / Carbon 42 (2004) 3251–3272
Kish graphite as a magnetic field sensor Yutaka Kaburagi *, Yuichiro Asano, Yoshihiro Hishiyama Faculty of Engineering, Musashi Institute of Technology, 1-28-1, Tamazutsumi, Setagaya-ku, Tokyo 158-8557, Japan Received 14 May 2004; accepted 6 August 2004
Keywords: A. Carbon films, Highly oriented graphite; D. Defects, Galvanomagnetic properties
The materials showing notable magnetoresistance effect (MRE) and Hall effect have been applied to magnetic field sensors. As materials for the static field sensor due to MRE, semiconductor compounds, such as InSb, InAs and GaAs, and ferromagnetic metals, such as Ni–Co and Ni–Fe, have been used. The semiconductor sensors exhibit positive MRE and can measure magnetic fields between 106 and 102 T. The ferromagnetic sensors exhibit negative MRE and can only measure low magnetic fields down to 1010 from 102 T. On the other hand, highly oriented graphite materials, such as high quality natural graphite (NG), Kish graphite (KG) and highly oriented pyrolytic graphite (HOPG) are known to exhibit large positive magnetoresistance [1–4]. However, such graphite materials have not been used as MRE sensors. Transverse magnetoresistance Dq/q0 is expressed as D q/q0 = [q(B) q(0)]/q(0), where q(0) and q(B) are the resistivity of the sensor at zero-magnetic field and that at a magnetic field B applied perpendicular to the current flowing in the sensor, respectively. Graphite is a semimetal and has almost equal densities of electrons and holes. For such material, Dq/q0 is approximately proportional to the square of mean mobility pffiffiffiffiffiffiffiffiffi lð¼ le lh Þ, i.e., Dq/q0 ffi l2B2 = lelhB2, where le is the mean electron mobility and lh is the mean hole mobility [1,3–5]. le and lh are very sensitive to the presence of lattice defects and impurities, and decrease with increasing their densities in the sensor. Therefore, the magnitude of Dq/q0 depends on the crystallinity of graphite sample, and high quality graphite with very low defect density shows very large Dq/q0. KG is a graphite flake precipitated from molten iron at high temperatures [4,6]. When the KG flakes are produced at a very high temperature as high as that at
*
Corresponding author. Tel./fax: +81 3 5707 2204. E-mail addresses: [email protected] (Y. Kaburagi).
0008-6223/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2004.08.006
which iron evaporates and purified carefully, they have much higher crystallinity than that of HOPG and exhibit a nature similar to single crystal graphite [2,6–8]. A high quality KG sample, therefore, exhibits much larger Dq/q0 than that of HOPG, and comparable to that of a high quality NG [1,4,8,9]. In the present study, a static magnetic field sensor was tried to prepare using the KG flakes showing Dq/q0 higher than 100% at a field of 1 T at room temperature. Dq/q0 can be denoted as Dq/q0 = DV/V, where V is the voltage of the sensor at zero-field and DV is the voltage change induced by magnetic field. Therefore, the magnitude of B can be represented as that of DV with a constant electrical current supplied to the sensor. For a rod-shaped sensor of rectangular cross-section with the width w and thickness t, DV is given by DV = [q(B) q(0)](I‘/wt), where ‘ is the distance between the two electrodes for the voltage measurement and I is the electric current. DV can be obtained as a larger value with a larger ‘ and smaller w and t at a constant I. In the case of KG samples, ‘ is limited because they are small flakes and I should not be large for suppressing Joule heating which rises to fluctuation of DV. Therefore, reducing w and t is important factor to increase DV, but a careful treatment is necessary to reduce w and t, i.e. without increase of the defect density of the sample. The sizes of the KG samples prepared in the present experiment were those with 2 mm width and lower than 0.3 mm in thickness. The original KG flakes have irregular shapes, with their approximate sizes of 15 mm · 8 mm · 0.5 mm. Each of the flakes was sandwiched between artificial graphite plates and glued to them. The flakes sandwiched by artificial graphite plates were cut into segments in rectangular shape with 6 mm · 2 mm in size, and then KG specimens were separated by dipping the segments into the solvent for the glue. Thin samples with flat surfaces and various values in thickness were prepared by peeling off carefully the both surfaces of the rectangular KG specimens using adhesive tape. The KG samples thus
Letters to the Editor / Carbon 42 (2004) 3251–3272
3267
obtained were 0.256, 0.146, 0.080, 0.050 and 0.040 mm in thickness and are denoted hereafter as KG followed by the thickness, such as KG-0.256. For each sample, V was measured at zero-magnetic field, and then DV was measured at a magnetic field after V had been offset electrically. The field dependence of DV (i.e., D V/V = Dq/q0 in a constant V) was measured at fields up to 1.0 T in the field perpendicular to the sample surface at room temperature (about 20 C). For the measurements, four-terminal-method was performed and a constant current of 50 mA flowed. The DV measurement was also made in the reverse direction of the field to eliminate Hall voltage. However, the Hall voltage appeared is very small and negligible in comparison
with the voltage due to MRE. In addition, to obtain larger DV in the low field region, a commercially available neodymium disk magnet with 4 mm / and 2 mm thick was used as a bias field magnet because D q/q0 is proportional to B2 in the low magnetic field region as mentioned above. Namely, each KG sensor was equipped with the bias magnet after above measurements. The bias field applied to each sensor was 130 mT. It should be noted that the bias field direction has to coincide with the field direction to be measured. We obtain wrong value of DV if the bias field is opposite in direction to the measuring field. In Fig. 1, the field dependence of Dq/q0 at fields up to 1 T at room temperature (a) and the low field behavior
Fig. 1. Field dependence of transverse magnetoresistance at room temperature at fields up to 1.0 T (a) and 0.2 T (b) for KG samples with various values in thickness and without bias magnet.
Fig. 2. Field dependence of DV at fields up to 1 T (a) and 0.2 T (b) at room temperature for KG samples with various values in thickness without bias magnet.
3268
Letters to the Editor / Carbon 42 (2004) 3251–3272
(b) are shown for the KG samples with various values in thickness. These KG samples can measure much higher fields than 1 T according to Dq/q0 ffi l2B2. KG-0.256 is a relatively thick sample obtained by just removing irregular surfaces on both sides of one of the rectangular KG specimens and exhibits a positive Dq/q0 larger than 100% at a field of 1 T, indicating highly crystallized nature. Dq/q0 changes randomly with sample thickness as shown in Fig. 1. Defects may be introduced by peeling off the sample surfaces. The defect density or crystallinity of sample does not depend on the sample thickness, rather it depends on the peeling off condition, crystallinity of the original KG specimen and also crystallinity of the remained part after peeling off because KG flakes are polycrystals and aggregates of oriented large crystal grains [10]. Thin samples of KG-0.050 and KG-0.040 exhibit larger Dq/q0 than others, which means that defects are not introduced remarkably by reducing their thickness and the parts with good crystal condition are remained. For all samples, the fluctuation of Dq/q0 was slight on the temperature change of several degrees around 20 C, and the reproducibility of Dq/q0 measured at a constant magnetic field on different days was good, less than 2%. Fig. 2 shows the field dependence of DV at fields up to 1 T at room temperature (a) and the low field behavior (b) for the KG samples with various values in thickness. At a constant magnetic field, DV increases with decreasing sample thickness, and is 18 lV at a field of 0.1 T for KG-0.040. If the defect density increases with reducing the sample thickness t, Dq/q0 decreases, i.e., q(B) q(0) decreases. However, DV increases when the rate of decrease in t surpasses that in q(B) q(0) be-
Fig. 4. DV versus magnetic field plots at room temperature for KG samples equipped with bias magnet.
cause of DV = [q(B) q(0)](I‘/wt) as written above. For each sample, D V, as well as D q/q0 shown in Fig. 1, is roughly proportional to B2. Fig. 3 shows the field dependence of DV at fields up to 0.2 T at room temperature for the KG samples equipped with the bias magnet. At a constant magnetic field, DV increases remarkably with increasing B because of the bias magnet equipped to the samples, and each curve is closer to a straight line than that without bias magnet. In Fig. 4, the DV versus B plots are shown. KG-0.040 equipped with the bias magnet, being the most sensitive field sensor obtained in this study, exhibits a sensitivity of about 0.8 lV/mT around 0.1 T with a constant current of 50 mA at room temperature. The sensitivity enables us to measure magnetic field down to the order of mT using a conventional volt-meter. There is a possibility to measure magnetic field in the order of 104 T by reducing further the thickness and width of the KG sample, though the sensitivity will still be inferior to the semiconductor and ferromagnetic sensors.
References
Fig. 3. Field dependence of DV at fields up to 0.2 T at room temperature for KG samples equipped with bias magnet.
[1] Soule DE. Magnetic field dependence of the Hall effect and magnetoresistance in graphite single crystals. Phys Rev 1958;112(3):698–707. [2] Hishiyama Y, Ono A, Tsuzuku T, Takezawa T. Galvanomagnetic properties of well-oriented graphite in relation to the structural imperfections. Japn J Appl Phys 1972;11(7):958–63. [3] Kreps LW, Woollam JA. Temperature and field dependencies of galvanomagnetic effects in graphite. Carbon 1977;15: 403–09. [4] Kaburagi Y. Magnetoresistance of highly oriented graphite at 77 K and 4.2 K. J Phys C: Solid State Phys 1982;15:5425–40.
Letters to the Editor / Carbon 42 (2004) 3251–3272 [5] Dresselhaus MS, Dresselhaus G, Sugihara K, Spain IL, Goldberg HA. Graphite fibers and filaments. Berlin: Springer-Verlag; 1988. p. 202–216. [6] Inagaki M. New Carbons—Control of structure and functions. Amsterdam: Elsevier; 2000. p. 30–39. [7] Yoshida A, Hishiyama Y. Electron channeling effect on highly oriented graphites—Size evaluation and oriented mapping of crystals. J Mater Res 1992;7(6):1400–5. [8] Kaburagi Y, Yoshida A, Hishiyama Y. Microtexture of highly crystallized graphite as studied by galvanomagnetic properties
3269
and electron channeling contrast effect. J Mater Res 1996;11(3): 769–78. [9] Hishiyama Y, Kaburagi Y. Relationship between residual resistivity ratio qRT/q4.2 K and transverse magnetoresistance in highly oriented graphite. Tanso 1988;13:394–9. in Japanese. [10] Kaburagi Y, Yoshida A, Hishiyama Y. Microtexture of highly crystallized graphite as studied by galvanomagnetic properties and electron channeling contrast effect. J Mater Res 1996;11(3): 769–78.
Filling of multi-walled carbon nanotubes with tin(IV) oxide Liping Zhao, Lian Gao
*
The State Key Laboratory of High Performance Ceramics and Superfine Microstructure, Shanghai, Institute of Ceramics, Chinese Academy of Sciences, 1295 Ding Xi Road, Shanghai 200050, PR China Received 1 March 2004; accepted 11 August 2004
Keywords: A. Carbon nanotubes; B. Chemical treatment
Multi-walled carbon nanotubes (MWNTs) have attracted much interest since their discovery owing to their high mechanical strength and electrical properties [1,2]. Another fascinating aspect of carbon nanotubes is their cavities. Only a few nanometers in diameter and a few micrometers in length, such a cavity should allow for interesting nanoscale experiments, if it can be filled with other materials in a systematic way. Theoretical studies suggest that the introduction of foreign material into the inner cavities of MWNTs may enhance or modify the properties of the resulting composite materials [3,4]. The procedure to fill carbon nanotubes can be classified in two groups: (a) the chemical method [5–7], and (b) the physical method [8–10]. In the physical method, no solvent is used in the process and opened tubes are directly immersed in the molten material whereby the liquid is driven into the tube by capillary forces [7,10,11]. Thus, the physical method is more restrictive in the material choice and the ability of such materials to be so incorporated depends on three characteristics [12]: (i) they must ÔwetÕ the capillaries and have surface tensions below a threshold value in the range 100– 200 mN m1 [13]; (ii) their overall melting temperature must be low enough to preclude thermal damage to the MWNT; and (iii) they must not attack MWNTs *
Corresponding author. Tel.: +86 24 2397 1863; fax: +86 24 2389 1320 E-mail address: [email protected] (L. Gao). 0008-6223/$ - see front matter 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.carbon.2004.08.009
chemically. The wet chemical method consists of the opening of nanotube ends and the inclusion of salt solution into the opened tubes. The main advantage of the wet chemical approach is its flexibility and the level of experimental control, so that a wider variety of materials can be introduced into carbon nanotubes than the physical method. SnO2 is one of the most important and useful semiconductor materials with a large band gap, which has been extensively used for gas sensors [14], transparent conducting electrodes [15], transistors, and solar cells [16]. Sloan et al. [17] reported a two-step wet chemical method using SnCl2 as a precursor for the filling of MWNTs with SnO nanoparticles in high yield, involving (a) opening and (b) filling. Unfortunately, they still need to dissolve SnCl2 in hot concentrated HCl solution first and the solid-state SnO is thermodynamically unstable with respect to b-Sn and SnO2. The decomposition reaction of 2SnO ! SnO2 + Sn was reported to occur starting at a temperature as low as 370 C [18]. Here we use a weak acidic SnCl2 solution as precursor for the filling of MWNTs with SnO2 nanocrystals. The addition of a small amount of acid is only for reducing the hydrolysis of SnCl2. MWNTs prepared by the catalytic decomposition of CH4 were kindly provided by Shenzhen Nanotech Port Ltd. Co. (Shengzhen, China) and used as received. The inner and outer diameters of the MWNTs measured by transmission electron microscopy (TEM,