Attachment of Single CdSe Nanocrystals to Individual Single-Walled

Eastman Kodak Company, Rochester, New York 14650. Received September 17, 2002. ABSTRACT. Single amine functionalized CdSe quantum dots were ...
0 downloads 0 Views 254KB Size
NANO LETTERS

Attachment of Single CdSe Nanocrystals to Individual Single-Walled Carbon Nanotubes

2002 Vol. 2, No. 11 1253-1258

Joanne M. Haremza, Megan A. Hahn, and Todd D. Krauss* Department of Chemistry, UniVersity of Rochester, Rochester, New York 14627

Samuel Chen and Joaquin Calcines Eastman Kodak Company, Rochester, New York 14650 Received September 17, 2002

ABSTRACT Single amine functionalized CdSe quantum dots were coupled to individual acid-chloride modified single-walled carbon nanotubes via amide bond formation. The quantum dot−nanotube conjugates were characterized by atomic force and transmission electron microscopies. For single-walled nanotubes with lengths of < 200 nm, binding of quantum dots occurred at the nanotube ends, whereas longer nanotubes showed evidence of sidewall attachment. The number of bound quantum dots per nanotube can relatively be controlled by varying the quantum dot/nanotube mass ratio. These quantum dot−nanotube complexes have the potential to serve as a basis for future nanoelectronic devices.

Single-walled carbon nanotubes (SWNTs) have been well studied in recent years due to their interesting physical properties.1,2 In particular, the unique electrical transport properties of carbon nanotubes suggest that they will have many advantages over conventional materials for use as superior nanometer scale wires. For example, recent studies have shown that metallic SWNTs can act as ballistic conductors3,4 and can carry current densities that are orders of magnitude larger than copper wires of the same diameter.5 Thus, as the demand increases for smaller and faster electronic devices, SWNTs may eventually play an important role in this miniaturization process. Taking advantage of their unmatched conductive properties and nanometer size, possible applications for SWNTs include novel electronic devices,6 such as transistors,7 batteries,8 and molecular computers.9 Colloidal CdSe semiconductor quantum dots (QDs) are highly emissive, spherical, inorganic particles typically a few nanometers in diameter. Due to the effects of quantum confinement on the photoexcited exciton, these materials have electronic properties that depend on the size of the particle. Semiconductor QDs show great promise for use in future photonics applications, such as solar cells10 and light emitting devices,11-13 because the wavelength response of the device can be tuned by simply varying the size of the quantum dot. While semiconductor QDs have great potential as future photonic materials, in practice a number of obstacles * Corresponding author. Phone: (585) 275-5093. Fax: (585) 506-0205. E-mail: [email protected] 10.1021/nl025799m CCC: $22.00 Published on Web 10/22/2002

© 2002 American Chemical Society

must be overcome in order for them to supplement current technology. For example, the quantum efficiencies of hybrid conducting polymer-QD light emitting devices are on the order of 0.25%.12 Hybrid CdSe nanorod-polymer solar cells perform slightly better, but their power conversion efficiencies are less than 2%,10 which is an order of magnitude lower than conventional inorganic solar cells. The relative inefficiency of these QD devices is largely due to the difficulty in effectively transferring charge carriers into (LED) or away from (solar cell) the CdSe nanoparticles composing the device. To raise the efficiency of QD devices to a practical level, techniques that facilitate charge transport directly to the quantum dot must be developed. For instance, attachment of conducting SWNTs to CdSe QDs would place a metallic wire in direct chemical contact with the quantum dot surface. The metallic SWNT could then promote direct charge transport and efficient charge transfer to the quantum dot; this system has the potential to significantly increase device efficiency. Development of novel nanometer-scale integrated systems incorporating SWNTs will likely require covalent attachment of the SWNT to another nanometer-sized material in a reliable and robust manner. However, binding of individual SWNTs to materials such as quantum dots has received little attention, in part due to the difficulty in processing the raw SWNT material. Recent attempts at connecting carbon nanotubes to nanoparticles include attachment of SWNTs to gold nanoparticles14 and connection of SWNT bundles

Scheme 1.

Formation of QD-SWNT Complexes via a Chemical Coupling Procedurea

a QD attachment can occur at cut nanotube ends or at defects along the sidewall. TOPO on the initial QDs has been omitted for clarity. Figure is not to scale.

with TiO2 and CdSe quantum dots.15 However, the methods used in the latter case contain undesirable side-reactions that should predominantly yield large clusters of QDs and nanotubes, which are unsuitable for use in any practical nanoelectronics application. In this letter, we will demonstrate the coupling of individual SWNTs to single semiconductor quantum dots. CdSe core and CdSe/ZnS core/shell quantum dots were attached to SWNTs via an amide coupling scheme. Attachment of QDs to both thin ropes and individual SWNTs was verified with atomic force microscopy (AFM) and transmission electron microscopy (TEM). By varying the mass ratio of SWNTs to CdSe QDs, the relative number of quantum dots per nanotube could be controlled. This work is an initial step toward obtaining light emissive devices on the single quantum dot level. Laser-oven synthesized SWNTs were purchased from Carbon Nanotechnologies Incorporated and Tubes@Rice. The SWNTs were purified and shortened with aqua regia according to previously reported methods, with treatment times ranging from 6 to 24 h.14 These shortened SWNTs were then functionalized with thionyl chloride according to methods previously discussed in the literature.16 CdSe quantum dots were synthesized in coordinating solvents using variations of the methods developed by Murray et al.17 and Qu et al.18 Quantum dot core diameters ranged from 2.8 to 4.3 nm as determined from optical absorption spectroscopy and atomic force microscopy. Certain CdSe QD cores were further capped with a semiconductor shell (ZnS) according to literature methods.19,20 Directly from the synthesis, CdSe QDs and CdSe/ZnS core/shell QDs were passivated with trioctylphosphine oxide (TOPO). The TOPO on the surface of the QD was exchanged for 2-aminoethanethiol (AET) according to the following procedure. TOPO was “washed” from the QD surface and solution by repeated cycles of precipitation with a butanol/ methanol mixture, centrifugation, and redispersion of the 1254

precipitate into hexane. The precipitate after the final washing cycle was redispersed in a neat solution of 2-aminoethanethiol‚ HCl and heated at 85 °C under nitrogen for 6 h to 2 days. The resulting (alcohol soluble) QD solution was rinsed several times with ethanol using Millipore Centriplus YM50 centrifugal filters (MWCO 50 000) to remove any unreacted ligand. The procedure for coupling the functionalized SWNTs and QDs is outlined in Scheme 1. Acid-chloride functionalized SWNTs were sonicated in ethanol for 30 min. The amine functionalized QDs were added to the SWNTs and refluxed at 90 °C under nitrogen for 4 days. The suspension was then cooled to room temperature, centrifuged for 30 min, and rinsed thrice with ethanol using several cycles of centrifugation and decantation. The coupled QD-SWNTs were filtered through a 3.0 µm Teflon membrane filter and resuspended in ethanol. The ratio of QDs to SWNTs was varied from 500:10 mg QD/SWNT to 50:140 mg QD/SWNT. For control experiments, dihydrolipoic acid (DHLA) was prepared from thioctic acid according to published methods21 and exchanged for the TOPO on the surface of the CdSe QD. To effect this exchange, QDs were stripped of TOPO via a repeated washing procedure using precipitation and centrifugation, added to 200 µL DHLA, and stirred overnight at 65 °C under nitrogen.22 Upon cooling, 10 mL of ethanol was added, and the solution was filtered and rinsed three times with ethanol using YM-50 centrifugal filters. DHLA functionalized QDs and acid-cut SWNTs were reacted under conditions identical to those employed during the amine functionalized QD coupling to the acid-chloride functionalized SWNTs. UV-vis absorption spectra were obtained on Perkin-Elmer Lambda 19 and Lambda 900 spectrometers. AFM images were obtained using a Digital Instruments Nanoscope IIIa. TEM images were obtained from a JEOL JEM-2000FX electron microscope at 200 kV. Nano Lett., Vol. 2, No. 11, 2002

Figure 1. Absorption spectra of 2-aminoethanethiol functionalized CdSe/ZnS core/shell QDs in ethanol (dashed line) and assynthesized QDs in hexane (solid line).

As-synthesized CdSe quantum dots are capped with TOPO ligands, and thus are hydrophobic. To obtain QDs that can bind to the acid-chloride functionalized SWNTs, the TOPO must be replaced with a suitable hydrophilic ligand. We chose to derivatize the CdSe QDs with amines, which should attach covalently to the acid-chloride functionalized SWNTs though the formation of an amide bond. (See Scheme 1.) After exchange of TOPO for AET, the CdSe QDs were stable in alcohols, such as ethanol, indicating a successful replacement of the TOPO with the hydrophilic ligand. Figure 1 shows typical absorption spectra of CdSe/ZnS core/shell QDs before and after functionalization with AET. Quantum dots capped with TOPO or AET both display a peak at 524 nm due to exciton absorption. The fact that the first exciton state does not blue shift upon ligand substitution suggests that substantial surface oxidation did not occur. Additional evidence that the QDs are capped with AET comes from infrared spectroscopy, where the presence of a broad amine peak at 3400 cm-1 indicates the desired ligand exchange occurred on the quantum dot surface. (See Supporting Information.) Careful and repeated rinsing was employed to remove any unreacted ligand and to ensure that the amine peaks observed in the FT-IR spectrum correspond to amines on the CdSe QD surface. The ligand exchange with AET failed approximately 30% of the time. After heating overnight, sometimes the colloidal solution would become colorless, indicating that the quantum dots had decomposed. We believe this failure is related to the presence of tetradecylphosphonic acid (TDPA), which was incorporated in certain CdSe QD syntheses.18 TDPA binds very strongly to Cd and was shown to be present on the QD surface after synthesis.23 The decomposition never occurred with quantum dots synthesized using Cd(CH3)2, where only TOPO is a surface ligand,17 or with QDs made with CdO and no TDPA present in the reaction mixture.18 Nano Lett., Vol. 2, No. 11, 2002

Figure 2. AFM topographic image of a small bundle of SWNTs on highly ordered pyrolytic graphite (HOPG). The height of the bundle is ∼3 nm. The sample was prepared by spin coating 10 drops (∼20 µL per drop) of a nanotube suspension in toluene (O.D. ∼ 0.1 at 2200 nm) onto the HOPG substrate after 30 min of sonication.

Purification of carbon nanotubes with aqua regia (i.e., acid cutting) results in clean sidewalls and eliminates any amorphous carbon particles present as a result of the synthesis. (See Figure 2.) Acid cutting was necessary to ensure that the particles attached to the SWNTs after the coupling reaction were due to attachment of CdSe QDs and not leftover carbonaceous particles. In agreement with previous studies,14 we found the SWNTs were shortened and the large bundles became exfoliated due to the acid cutting procedure. Typical AFM images depicting the attachment of single CdSe QDs to SWNTs are shown in Figure 3. Acid-chloride functionalized SWNTs that have been reacted with amine functionalized CdSe QDs have nanoparticles connected to the sidewalls and attached to the nanotube ends. As seen in Figure 2, the AFM image of a nanotube before functionalization is devoid of any particle-like features, thus suggesting that functionalization of the SWNTs with QDs was successful. The inset to Figure 3a depicts a cross-section analysis of a typical QD-SWNT complex. The height of the nanotube plus the height of the quantum dot equals the total height at the QD-SWNT junction. In this case, the nanotube height is 1.2 nm, suggesting that this image is of a single nanotube and not a small bundle. The total junction height is 6.1 nm, and thus we determine the height of the particle in the image to be 4.9 nm. This measurement agrees with the size of the CdSe QD as obtained from optical spectroscopy17 and from independent AFM measurements. Similar measurements of QD-SWNT junctions provide additional evidence that the particles attached to the SWNTs are indeed individual CdSe QDs. 1255

Figure 4. (a) TEM images of the QD-SWNT complexes. The samples were prepared by placing 2 drops (∼20 µL per drop) of the CdSe QD-SWNT suspension in ethanol onto ashed lacey carbon grids. Images were obtained at 200 kV. Individual SWNTs can be detected from within the small bundles, and single QDs (dark spots) are bound at various points along the nanotube walls. (b) TEM images from the control experiment, obtained in a manner similar to that described above. These images show only SWNTs with no sign of QDs bound to the nanotubes.

Figure 3. (a) AFM image of a single carbon nanotube with one CdSe QD attached at either end. The QDs appear slightly elongated along the vertical direction due to piezo drift. The inset shows cross sections along the lines in (a). (b) Additional AFM image of a nanotube with attached QDs along its sidewall. QD-SWNT AFM samples were prepared by spin coating 10 drops (∼20 µL per drop) of the suspension in ethanol (O.D. ∼ 0.1 at 2200 nm) onto a mica substrate after 30 min of sonication.

Unreacted individual quantum dots and QD aggregates are also typically observed in the AFM images. The frequency of these observations is a function of the QD amount used for the coupling; larger amounts of QDs typically result in more free particles present in the images. While the majority of unreacted QDs should have been removed during the filtration process, this procedure retains some free particles in the final suspension. It is possible that the QD-SWNT complexes observed using AFM are not due to covalent attachment of individual QDs to SWNTs but rather to accidental resting of free QDs along SWNTs due to the spin coating process. Therefore, transmission electron microscopy was also used to confirm that individual CdSe QDs were bound to SWNTs. As shown in Figure 4a, thin SWNT bundles containing as little as four 1256

nanotubes are observed to have single QDs bound at various points along their sidewalls. The diameters of the dark, strongly scattering particles in the image correspond to the same CdSe diameter as determined from optical absorption data and from AFM measurements, thus further indicating covalent attachment of single QDs. It is important to note that both AFM and TEM indicate that most of the observed quantum dot binding to SWNTs occurs with single QDs. To prove that the CdSe QDs and the SWNTs are covalently attached through an organic linkage and not through nonspecific binding interactions, shortened SWNTs containing carboxylic acid groups were reacted with CdSe/ ZnS QDs functionalized with DHLA. For SWNTs and QDs fuctionalized in this manner, no covalent attachment is expected. As shown in Figure 4b, TEM images of SWNTs reacted with DHLA functionalized quantum dots do not contain any particle-like features; the individual sidewalls of each SWNT in the bundles can be seen clearly and are featureless. Thus, we conclude that nonspecific QD attachment did not occur and that the attachment of the QDs functionalized with AET to individual SWNTs occurred through amide bond formation. We repeated the QD-SWNT coupling reactions with nanotubes synthesized using different methods, with quantum dots having core diameters that ranged from 2.8 to 4.3 nm, and with CdSe core and CdSe/ZnS core/shell QDs. In the absence of TDPA, the coupling results were not affected by the particular method used to synthesize the CdSe QDs or the presence of a semiconductor (ZnS) shell around the quantum dot core. The QD/SWNT mass ratio was varied from 9:1 to 0.37:1. For a heavily quantum dot rich procedure, we noted CdSe Nano Lett., Vol. 2, No. 11, 2002

QD attachment to a significant portion of the nanotube sidewalls, indicating that the acid cutting process attacks not only the nanotube ends but also a remarkable percentage of the sidewalls. This observation is consistent with Raman studies showing that the D-band Raman intensity (the measure of sp2 disorder in a nanotube) increases significantly upon acid purification of the nanotubes.24,25 This result also agrees with studies that have shown a very high percentage of acidic carbon sites on purified SWNT material.26,27 As the QD/SWNT ratio decreased, the number of attached QDs per nanotube bundle decreased, as expected. When the QD/ SWNT ratio was < 1, we saw rare instances (less than 1% of the time) of individual QDs connecting two single nanotubes or two small ropes of nanotubes. This configuration of an individual quantum dot connecting two nanotubes is the precursor for an emissive type of single quantum dot nanodevice. In this scenario, positive and negative charge carriers would be transported separately across each conducting SWNT to the QD, where they would recombine and emit light at a wavelength corresponding to the QD energy gap. Unfortunately, this particular form of QD-SWNT conjugate was rarely observed. The probability of attaching QDs primarily at the nanotube ends, preferred for simple device configurations, is a function of the SWNT length. For SWNTs cut with acid for 22 h, the majority of nanotubes were very short segments on the order of 150-250 nm long, as determined by AFM. After formation of the CdSe quantum dot-SWNT complex, the CdSe QDs are attached primarily at the nanotube ends. (See Figure 3a.) However, for these very short nanotubes, the likelihood of sidewall attachment versus end binding is small, and these results are somewhat expected. Also, these nanotubes are currently too short to be of practical use for connecting to the macroscopic world. For longer nanotube bundles, QDs were attached primarily along the tube sidewalls, and selective functionalization of nanotube ends was rarely observed. To have binding solely at the ends of the SWNTs, they need to be short enough to ensure there will be no acid groups on the sidewalls, which is unlikely for nanotubes longer than ∼200 nm. Using both AFM and TEM to characterize QD-SWNT conjugates is important, since these experiments can provide complementary information. For example, the interesting instances where a single QD connects the ends of two nanotubes was seen only with AFM; this phenomenon primarily occurred for short nanotubes, which fall through the holes in a lacey carbon TEM grid. TEM can better distinguish whether QDs are attached to individual nanotubes or to a nanotube bundle, since the number of individual SWNTs in a bundle can easily be seen in an image. Figure 5 illustrates the absorption spectra of the SWNTs before and after the coupling reaction with the CdSe QDs. Three electronic transitions are clearly observed in the spectrum of the nanotubes. The two peaks found at lower energy correspond to the two lowest optically allowed transitions of semiconducting SWNTs. The third peak is the lowest optically allowed transition for metallic tubes.28,29 Comparison of the SWNT absorption before and after the Nano Lett., Vol. 2, No. 11, 2002

Figure 5. Typical absorption spectra of shortened SWNTs (dashed line) and a QD-SWNT complex (solid line). The samples were prepared on a quartz coverslip through evaporation of the solvent. Both spectral lines show the characteristic peaks expected from SWNTs.

coupling shows that the SWNT electronic structure is not altered as a result of the coupling to CdSe QDs. Similar spectra were obtained for the entire range of QD/SWNT ratios studied (9:1 to 0.37:1 by mass). Features corresponding to CdSe QD absorption were never observed (Figure 1), indicating that the relative amount of QDs that successfully bound to the SWNTs was small. It has been previously reported that the covalent attachment of CdSe QDs to SWNTs results in a charge transfer between the two materials, as observed through a change in the absorption spectra of the SWNTs after the reaction.15 However, as demonstrated in the absorption spectra of Figure 5, we have found no evidence of any charge exchange or electronic state coupling between the nanotubes and the QDs, at a level of up to approximately 30 QDs per micron of tube length. Modified absorption spectra caused by supposed coupling of SWNTs and CdSe QDs implies a significant electronic wave function overlap between the QD and the SWNT. However, for a CdSe QD, the electron and hole wave functions are well confined to the quantum dot interior;20 as a result, they are not expected to significantly couple to any external electronic states. Furthermore, a CdSe QD with a diameter of ∼3.5 nm has an electron affinity of 4.7 eV and an ionization potential of 6.8 eV.12 Since these values are obtained from the values of bulk CdSe modified only by quantum confinement theory,30 other CdSe QDs with diameters of 3.5 ( 1 nm should have electron affinity and ionization potential values differing by less than 0.15 eV.12 In addition, graphite has a work function of 5.0 eV.31 If the work function of a nanotube is also assumed to be 5.0 eV,32 then the nanotube Fermi level lies in the HOMO-LUMO gap of the CdSe QD, and no charge transfer is expected. However, it is possible that the discrepancy between our results and that of Banerjee and Wong15 could be due to 1257

other factors, such as differences in SWNT preparation method or in the specific ligand used to couple the QD to the SWNT. Our results have direct implications for self-assembly of carbon nanotube nanodevices using chemical attachment strategies. Carboxylic acid groups present on the nanotube sidewalls, as well as the ends, lead to uncontrolled amide bond coupling. This nature of coupling may be acceptable for some applications, such as chemical or biological sensors,33,34 but is not practical for electronic devices where the positioning of the nanotube is critical. Also, the harsh acidic processing of SWNTs could break the π bonding symmetry of the sp2 hybridization because of the many defects in the sidewalls, thus mitigating their unique electronic properties. In order for SWNT devices to surpass the performance of present devices, the electronic structure of the nanotubes should remain intact. In conclusion, we present a method for coupling individual SWNTs to single CdSe QDs. For relatively long nanotubes (>200 nm), the majority of QDs attached to the SWNT sidewalls. For relatively short tubes (