NANO LETTERS

Filling Carbon Nanotubes with Particles

Byong M. Kim, Shizhi Qian, and Haim H. Bau*
Department of Mechanical Engineering & Applied Mechanics, UniVersity of PennsylVania, Philadelphia, PennsylVania 19104-6315
Received February 12, 2005; Revised Manuscript Received March 22, 2005

2005 Vol. 5, No. 5 873-878

ABSTRACT
The filling of carbon nanotubes (CNTs) with fluorescent particles was studied experimentally and theoretically. The fluorescent signals emitted by the particles were visible through the walls of the nanotubes, and the particles inside the tubes were observable with an electron microscope. Taking advantage of the template-grown carbon nanotubes transparency to fluorescent light, we measured the filling rate of the tubes with particles at room conditions. Liquids such as ethylene glycol, water, and ethylene glycol/water mixtures, laden with 50 nm diameter fluorescent particles, were brought into contact with 500 nm diameter CNTs. The liquid and the particles transport were observed, respectively, with optical and fluorescence microscopy. The CNTs were filled controllably with particles by the complementary action of capillary forces and the evaporation of the liquid. The experimental results were compared and favorably agreed with theoretical predictions. This is the first report on fluorescence studies of particle transport in carbon nanotubes.

The transport of simple and complex fluids in carbon nanotubes (CNTs) is of interest from both the fundamental science and the applications points of view.1-3 Carbon nanotubes are a convenient material with which to work for several reasons. First, carbon nanotubes can be fabricated with diameters ranging from a fraction of a nanometer to several hundreds of nanometers, allowing one to conduct experiments with various tube sizes. Second, the surface properties of the tubes can be modified thermally and/or chemically to facilitate behaviors ranging from hydrophilic to hydrophobic, allowing one to probe the effect of surface properties on the behavior of the liquids. Third, the walls of the tubes are sufficiently thin to be transparent to light4 and electrons,5-9 allowing one to observe and quantify events that take place inside the tube. Moreover, the tubes can contain high-pressure fluids and gases for an extended time, even in the vacuum environment of the electron microscope.5,6 The CNTs may also be useful as components in biosensors, minute chemical reactors, and drug delivery and nanofluidic systems, as well as constituent components in composite materials. Furthermore, the CNTs can be used as minimally intrusive nanopipets to probe biological cells.10 Indeed, recent studies indicate that these carbon tubes are benign to living cells.11 Recently, taking advantage of the optical transparency of template-grown CNTs with 15 nm thick walls,12,13 we studied the filling of the CNTs by capillary imbibition and condensation and the emptying of the tubes by evaporation.4 In this paper, we describe experiments focusing on filling the carbon nanotubes with a suspension. Liquid, laden with nominally
* Corresponding author. Tel: (215) 898-8363. Fax: (215) 573-6334. E-mail: bau@seas.upenn.edu. 10.1021/nl050278v CCC: $30.25 Published on Web 04/02/2005 2005 American Chemical Society

Figure 1. Schematic depiction of the filling process.

50-nm-diameter fluorescent nanoparticles, was brought into contact with one end of a CNT. The liquid and the particle transport were observed, respectively, with optical and fluorescence microscopy. The experimental observations were compared with theoretical predictions. In contrast to previous studies that focused on the translocation of particles in carbon nanotubes,3 our study addresses the process of particle filling and compaction and the formation of densely packed beds. The experimental setup is depicted schematically in Figure 1. The carbon nanotubes (CNTs) were prepared according to the procedure described elsewhere.7,14 Briefly, the CNTs were synthesized by chemical vapor deposition of carbon on the walls of pores in alumina membranes followed by dissolving the alumina in NaOH solution. The process

Figure 2. Filling of a CNT with particles. The column on the left depicts the fluorescent signal emitted by the particles through the tube wall. The column on the right is a cartoon of the filling process. The particle volume fraction in the drop is 0.004. The suspending liquid is a mixture of water and ethylene glycol.

yielded relatively straight, long tubes with open bores. The tubes ranged in length from 20 to 50 m and had an average diameter of approximately 500 nm and a wall thickness of 15 nm. The tube diameters were estimated from electron microscope images and ranged from 300 to 700 nm. The actual tube diameters may have been somewhat smaller due to tube deformation. The CNTs were then suspended in 2-propanol solution and placed on glass cover slips with the aid of dielectrophoresis.12,13,15 The 2-propanol was then dried away. Figure 2a is an optical image of an empty 29 m long CNT. A liquid microdroplet, laden with 50 nm nominal diameter fluorescent polystyrene beads, was placed at one end of the CNT with a glass micropipet with a 50 m diameter opening. The micropipet was attached to a micromanipulator. A length Xi of the CNT was immersed in the drop. The suspension was prepared by blending ethylene glycol (Fischer Scientific, E-178) with a suspension of fluorescent polystyrene beads (1% weight/volume, Spherotech, FP-00552-2) in deionized water. The fluorescent dye was incorporated (by the manufacturer) in the core of the particles. According to the manufacturer, the bead diameters ranged from 40 to 60 nm. Our electron microscope observations suggest that the beads were somewhat smaller. Two different blends were prepared. Blend A consisted of a 1:3 volume ratio of particle suspension and ethylene glycol with an approximate particle volume fraction of 0.002. Blend B consisted of a 1:1 ratio of particle suspension and ethylene glycol with an approximate particle volume fraction of 0.004. The suspension readily filled the CNT by capillary suction. The filling process was initially observed from above with an optical microscope, utilizing a dry objective optical lens at 1000 magnification. The portion of the tube occupied by the liquid was clearly visible as it had a darker color than the empty part of the tube. The filling process of the dilute suspension was similar to the filling process of pure liquids previously described by Kim et al.4 Once the feeding drop was removed, the liquid inside the tube evaporated readily, leaving the particles behind. During the evaporation process,
874

one end of the slug remained pinned to the downstream end of the tube while the other end of the slug receded. For additional details on liquid filling and evaporation, see Kim et al.4 Next, we repeated the filling observation using fluorescent light. The fluorescence was excited with a mercury lamp and a blue filter. The fluorescing particles were observed with a green filter. The left column of Figure 2b,c,d shows, respectively, the CNT at 10 s, 20 s, and 40 s after fluorescent light was first observed at the downstream end of the tube. The particle volumetric fraction in the drop was 0.004, and about Xi ) 3.7 m of the tube length was initially immersed in the drop. The images were taken with a digital camera, which recorded video images at the rate of 14 frames per second. Fluorescent light was first observed at the downstream end of the tube, and it propagated upstream until the entire length of the tube fluoresced. The location of the moving interface of the packed particles is indicated with vertical arrows. The large, bright blobs on the right-hand side of the images correspond to the droplet in contact with the tube inlet. Figure 2e shows the fluorescence image of the CNT, fully packed with particles, 5 s after removing the droplet from the tube inlet. The right column of the figure is a cartoon of the filling process. The portion of the tube closely packed with particles appears as a bright line segment. The rest of the tube that is filled with a low-density suspension appears dark. The experiments were repeated numerous times with good reproducibility. In most cases, the particle packing resulted in a continuous fluorescing line. Occasionally, we observed a dashed fluorescent line, most likely resulting from particles blocking the tube. Although the science of tube blocking in particulate flow is still in its infancy, available data suggest that such blocking should be a relatively low probability event in our experiments.16 The particle packing stopped when the droplet was separated from the tube inlet and resumed when the droplet was placed back in contact with the end of the tube. The filling process is likely due to the combined action of capillary forces and evaporation. The empty tube fills initially with low-density suspension by capillary action. At this stage, the particle concentration is too low for the fluorescent light to be visible through the carbon tube wall with our low sensitivity camera. During the filling process and thereafter, the front end of the liquid column evaporates, and the evaporating liquid is refurnished by capillary action. This process induces a continuous flow of the suspension from the drop into the tube and results in an increased particle concentration next to the evaporating surface. Eventually, a closely packed bed of particles forms at the downstream end of the tube. As the process continues, the length (Xf) of the packed bed increases. The packed array emits sufficiently high intensity fluorescent light that it was readily visible through the nanotube wall. As long as the tube was in contact with the drop, we did not observe any drying of the packed bed. Apparently, the liquid transport was mostly controlled
Nano Lett., Vol. 5, No. 5, 2005

Figure 3. TEM image of a particle-filled nanotube.

by evaporation, and the capillarity suction provided a sufficient inflow of liquid to make up for the evaporating fluid. These assertions were partially confirmed by direct visualization of the packed particles with a transmission electron microscope (TEM, JEOL 2010). We prepared TEM samples by first placing empty CNTs on a standard Cu-based TEM grid (Ted Pella, G600TT) with the aid of dielectrophoresis. From the trapped tubes, we selected the CNTs that were long enough to bridge across the 20 m 20 m square opening on the Cu grid. The CNTs were filled with particles using a process similar to the one depicted in Figure 1. Figure 3 shows an example of a TEM image featuring a portion of a nanoparticle-filled CNT. The presence of the nanoparticles in the CNT altered the tube transparency to the electron beam (200 kV), and the particles are clearly visible inside the tube. The results of several runs of the particle packing experiments are summarized in Figure 4. Figure 4 depicts the length (Xf) of the packed bed as a function of time (t). The length of the tubes used in these experiments ranged from 20 to 40 m, and the tubes had diameters of about 500 nm. The symbols and lines correspond, respectively, to experimental data and theoretical predictions. The mathematical model used for the theoretical predictions is described later in the paper. The solid and hollow symbols correspond, respectively, to suspensions having initial particle volumetric concentrations of 0 ) 0.002 (blend A) and 0.004 (blend B). The solid squares, circles, and triangles correspond, respectively, to Xi ) 1.3, 6.0, and 12 m. The hollow squares, circles, triangles, and diamonds correspond, respectively, to Xi ) 1.5, 3.7, 4.2, and 9.0 m. The filling rate depended on both the initial concentration of the suspension and the length of the tube immersed in the drop. When 0 ) 0.002, the average filling rate ranged from 0.1 to 0.4 m/s. When 0 ) 0.004, the average filling rate ranged from 0.3 to 1 m/s. The filling rate was nearly linearly proportional to 0. As the length (Xi) of the tube immersed in the drop increased, the filling rate decreased. Figure 5 depicts the average rate of the particle packing in m/s as a function of the length of the tube immersed in the drop (Xi in m). The solid and hollow symbols correspond, respectively, to suspensions having initial volumetric concentrations of 0 ) 0.002 (blend A) and 0.004 (blend
Nano Lett., Vol. 5, No. 5, 2005

Figure 4. Length of the packed bed (Xf) depicted as a function of time. The symbols and lines correspond, respectively, to experimental data and theoretical predictions. The solid and hollow symbols correspond, respectively, to particle volume fractions of 0 ) 0.002 (blend A) and 0 ) 0.004 (blend B). The solid squares, circles, and triangles correspond, respectively, to Xi ) 1.3, 6.0, and 12 m. The hollow squares, circles, triangles, and diamonds correspond, respectively, to Xi ) 1.5, 3.7, 4.2, and 9.0 m. The theoretical predictions correspond to various particle volume fractions of 0 ) 0.07% (green), 0.13% (magenta), 0.19% (blue), 0.40% (red), and 0.67% (black).

Figure 5. Average filling rate as a function of the length of the tube immersed in the drop. The solid and hollow symbols correspond, respectively, to particle volume fractions of 0 ) 0.002 and 0 ) 0.004. The solid lines are best-fit lines.

B). The values of the particle packing rates were calculated from the average slopes of curves similar to the ones depicted in Figure 4. The solid lines in Figure 5 are best-fit lines. The packing rate appears to depend nearly linearly on the length of the tube immersed in the drop. As the immersed length decreased, the filling rate increased. This can be attributed to the higher concentration of particles next to the surface of the drop due to the drops evaporation.17 We use a simple continuum, multiphase flow model to simulate our experiments. Consider a horizontal, cylindrical tube of radius R and length L initially filled with a
875

homogeneous suspension of particles of volume fraction 0. One end of the tube (x ) 0) connects to a reservoir with a well-mixed suspension of particles of volume fraction 0. At the other end (x ) L), the liquid evaporates. The x-coordinate spans the length of the tube. The suspension is described as a superposition of continua, and both the liquid and solid media are treated as incompressible viscous fluids.18 We assume that the transport processes in the suspension are controlled by the liquid evaporation at x ) L. In the model, we use the measured evaporation rate, so our theoretical predictions are independent of the tube diameter. The continuity equations for the solid and liquid phases are, respectively, + (vs) ) 0 t and (1 - ) + ((1 - )vf) ) 0 t (2) (1)

no permanent contact between the particles and e() ) 0. When > c, the particles are in contact and form a compressible packed bed, and stress forms due to particleparticle contact forces. The values of the parameters 0 > 0, K g 1, and c depend on the particles materials and shapes.20 Here, we select K ) 6, c ) 0.2 and 0 ) 180Pa.21 The resistance coefficient 18 f R() ) - 2 d V() (9)

where f is the viscosity of the fluid, d is the diameter of the particles, V() )

(1 - )n-2 (n > 2) for 0 e e max (10) 0 otherwise

and max 0.6 is the maximum volume fraction of the packed bed. We used n ) 4.65 (ref 22). Combining the continuity and momentum equations of the solid phase, we obtain + (q) ) A() t x x

where vs and vf are, respectively, the solid and fluid velocities, and is the volume fraction of the solid phase. By summing up equations 1 and 2, we obtain the equation q ) 0 for the volume-averaged velocity of the suspension q: ) (1 - )vf + vs ) - (1 - )u + vs (4) where (3)

(11)

A() )

(1 - )2 de() d R()

where u ) vs - vf is the slip velocity of the particles. Burger et al.19 derived simplified momentum equations for the solid and liquid phases in the form Isf e() ) x 1- and Isf p )x 1- In the above, p is the pressure, Isf ) -R()u is the solid-fluid interaction force, and (7)

(5)

The flow rate q is given by the evaporation rate of the liquid at the liquid-vapor interface (x ) L). The term q can be either estimated based on a simple diffusion model in the vapor phase4 or determined from experimental data. Our experiments indicate that the evaporation rate is nearly independent of time and the particle concentrations in the reservoir, and it is about 50 m/s. At the end of the tube that is connected to the reservoir, (0,t) ) 0 (12a)

(6)

At the left end of the tube (x ) L), we have the following boundary condition: q(1 - (L,t))2 de((L,t)) (L,t) )0 x R((L,t)) d(L,t)

(12b)

The simulation starts with the tube being filled with a homogeneous suspension: (x,0) ) 0 (13)

e() ) 0 c

{ (( )
0

- 1 if > c

if e c (8) We solved eq 11, subjected to the initial condition (eq 13) and boundary conditions (eq 12), using the explicit Kurganov-Tadmor central difference scheme.21,23,24 This scheme has the advantage of high resolution and low
Nano Lett., Vol. 5, No. 5, 2005

is the effective solid stress function. When e c, there is

876

Figure 6. The particle volume fraction (x,t) is depicted as a function of x at various times. Blend B (0 ) 0.004).

numerical (artificial) viscosity. Unfortunately, like other explicit schemes in conservation form, it requires small time steps to ensure numerical stability. The difficulty in solving eq 11 stems from the presence of a shock wave in the particle concentration.19,20 In the simulations, we considered a CNT of length L ) 40 m and particles of diameter d ) 50 nm. We carried out simulations for particles with other diameters such as 40 nm. Since the phenomenon was driven by the liquid evaporation, the simulation results were only mildly affected by changes in the particle diameter. For example, a reduction of 20% in the particle diameter resulted in 15% reduction in the filling rate. The viscosities of the blends were based on tabulated data for mixtures of water and ethylene glycol25 and were, respectively, 4.0 10-3Pas and 8.0 10-3Pas for the blends A and B. Since the flow was driven by evaporation, the results were not sensitive to the viscosity of the liquid. Since the volume fraction of the particles in the experiments was known only approximately, we adjusted it in the computations so that the simulation results matched the experimental data. Hence, 0 is an adjustable parameter in our simulations. Figure 6 depicts the particle concentration as a function of x at various times for blend B (0 ) 0.004). Witness the formation of a high density packed bed whose length increases with time. This behavior is consistent with the experimental observations. The fluorescent signal is visible through the tube wall only at high particle concentrations. As time progresses, the transition between high and low concentrations migrates upstream and the fluorescing line spreads from the downstream end of the tube toward the tube inlet. The predicted Xf is depicted as a function of time with various lines in Figure 4 for various initial particle volume concentrations of 0 ) 0.07%, 0.13%, 0.19%, 0.40%, and 0.67%. Qualitatively, the predictions are in a good agreement with the experimental observations (symbols). The length of the
Nano Lett., Vol. 5, No. 5, 2005

packed bed increases and the growth rate decreases slightly as time increases. Consistent with the experimental observations, as the particle concentration at the nanotube inlet increases, the growth rate of the packed column increases. When blend A is used, the experimental data (solid symbols) in Figure 4 is consistent with 0.19% e 0 e 0.075%. When blend B is used, the experimental data (open symbols) is consistent with volume fractions of 0.19% e 0 e 0.67%. In this investigation, we studied experimentally and theoretically the filling of carbon nanotubes with particles. Our results demonstrate that (1) particles can be transmitted through carbon nanotubes, (2) carbon nanotubes can be filled with particles, (3) the filling process at the length scales considered here can be reasonably predicted with continuum theories, (4) the thin walls of the carbon nanotubes are transparent to fluorescent light, and (5) the polymer-based nanoparticles inside the tubes can be observed with an electron microscope. The results of this work may be implemented in various ways. The ability to controllably fill nanotubes with particles allows one to modify the tube properties. In the work presented here, we endowed the tubes with the ability to emit fluorescent light by filling them with fluorescent particles. Similarly, the tubes can be filled with ferromagnetic particles,26 endowing them with magnetic properties. One can also envision using the tubes as containment vessels or nanoaquariums for biological studies of live macromolecules in the vacuum environment of the electron microscope, thus overcoming the limited spatial resolution of visible light. In the latter case, it would be necessary to seal the tubes with the specimens under investigation inside, a technology currently under development in our laboratory. Acknowledgment. This work was funded by the National Science Foundation through grant NSF-NIRT 0210579 and the Commonwealth of Pennsylvanias Nano Institute. Guzeliya Korneva from Drexel University prepared and supplied the CNTs used in this work. Davide Mattia from Drexel University operated the TEM. Helpful discussions with Prof. Y. Gogotsi from Drexel University were appreciated. References
(1) Supple, S.; Quirke, N. Phys. ReV. Lett. 2003, 90, 214501-1. (2) Bau, H. H.; Sinha, S; Kim, B.; Riegelman, M. (invited paper) In Proceedings of SPIE, Nanofabrication: Technologies, DeVices, and Applications; Lai, W. Y.-C., Pau, S., Lopez, O. D.; Eds.; SPIE: Philadelphia, October 25-28, 2004; Vol. 5592, pp 201-213. (3) Henriquez, R. R.; Ito, T.; Sun, L.; Crooks, R. M. Analyst 2004, 129, 478. (4) Kim, B. M.; Sinha, S.; Bau, H. H. Nano Lett. 2004, 4(11), 2203. (5) Gogotsi, Y.; Libera, J. A.; Guvenc-Yazicioglu A.; Megaridis, C. M. Appl. Phys. Lett. 2001, 79, 1021. (6) Megaridis, C. M.; Guvenc-Yazicioglu, A.; Libera, J. A.; Gogotsi, Y. Phys. Fluids 2002, 14, L5. (7) Rossi, M. P.; Ye, H.; Gogotsi, Y.; Babu, S.; Ndungu, P.; Bradley, J. C. Nano Lett. 2004, 4(5), 989. (8) Naguib, N.; Ye, H.; Gogotsi, Y.; Yazicioglu, A., G.; Megaridis, C., M.; Yoshimura, M. Nano Lett. 2004, 4(11), 2237. (9) Yarin, A. L.; Yazicioglu, A., G.; Megaridis, C. M. Appl. Phys. Lett. 2005, 86, 013109. (10) Kim, B. M.; Murray, T.; Bau, H. H. Nanotechnology, in revision. (11) Cherukuri, P.; Bachilo S. M.; Litovsky, S. H.; Weisman, R. B. J. Am. Chem. Soc. 2004, 126, 15638. 877

(12) Riegelman, M. A. Masters Thesis, University of Pennsylvania, 2004. (13) Riegelman, M.; Liu, H.; Bau, H. H. Trans. ASME, J. Fluid Eng., in press. (14) Bradley, J. C.; Babu, S.; Ndungu, P.; Nikitin, A.; Gogotsi, Y. Chemistry Preprint SerVer CPS: 030302 2003. (b) Che, G.; Lakshmi, B. B.; Martin, C. R.; Fisher, E. R.; Ruoff, R. S. Chem. Mater. 1998, 10, 260. (15) Riegelman, M. A.; Liu, H.; Evoy, S.; Bau, H. H. In Proceedings of NATO-ASI Nanoengineered Nanofibrous Materials; Guceri, S., Kutznetsov, V., Gogotsi, Y., Eds.; Kluwer: The Netherlands, 2004; pp 407-414. (16) Yamaguchi, E.; Adrian, R. J. Microchannel Blockage Phenomena 2004, private communication. (17) Routh, A., F.; Zimmerman, W., B. Chem. Eng. Sci. 2004, 59, 2961. (18) Drew, D. A.; Passman, S. L. Theory of Multicomponent Fluids; Springer-Verlag: New York, 1999.

(19) Burger, R.; Fjelde, K. K.; Hofler, K.; Karlsen, K. H. J. Eng. Math. 2001, 41, 167. (20) Burger, R.; Concha, F.; Karlsen, K. H. Chem. Eng. Sci. 2001, 56, 4537. (21) Berres, S.; Burger, R.; Karlsen, K. H.; Tory, E. M. SIAM J. Appl. Math. 2003, 64, 41. (22) Burger, R.; Karlsen, K. H.; Tory, E. M.; Wendland, W. L. Z. Angew. Math. Mech. 2002, 82, 699. (23) Kurganov, A.; Tadmor, E. J. Comput. Phys. 2000, 160, 241. (24) Qian, S.; Burger, R.; Bau, H. H. Chem. Eng. Sci. 2005, 60, 2585. (25) Handbook of Chemistry and Physics, 57th ed.; West, R. C., Ed.; CRC Press: Cleveland, OH, 1976-1977. (26) Korneva, G.; Ye, H., Gogotsi, Y.; Halverson, D., Friedman, G.; Bradley, J.-C.; Kornev, K. G., Nano Lett., 2005, 5, 879.

NL050278V

878

Nano Lett., Vol. 5, No. 5, 2005