Synthesis of Blue Luminescent Si Nanoparticles Using Atmospheric-Pressure Microdischarges NANO LETTERS 2005 Vol. 5, No. 3 537-541 R. Mohan Sankaran, Dean Holunga, Richard C. Flagan, and Konstantinos P. Giapis* DiVision of Chemistry and Chemical Engineering, California Institute of Technology, Pasadena, California 91125 Received December 1, 2004; Revised Manuscript Received January 23, 2005 ABSTRACT Silicon nanoparticles are synthesized from a mixture of argon/silane in a continuous flow atmospheric-pressure microdischarge reactor. Particles nucleate and grow to a few nanometers (1 3 nm) in diameter before their growth is abruptly terminated in the short residence time microreactor. Narrow size distributions are obtained as inferred from size classification and imaging. As-grown Si nanoparticles collected in solution exhibit room-temperature photoluminescence that peaks at 420 nm with a quantum efficiency of 30%; the emission is stable for months in ambient air. The promise of silicon-based optoelectronics 1 has spurred intense interest in Si nanoparticles (np-si) where direct band gap transitions have been reported as a result of quantum confinement. 2 Indeed, stable room-temperature photoluminescence, tunable in the range 700-350 nm, has been reported for np-si smaller in size than the excitonic radius for bulk Si ( 4 nm). 2-5 Most of the techniques used to synthesize np-si involve a capping agent for protection from uncontrolled oxidation which, however, may introduce surface recombination states that alter the emission characteristics. From this perspective, gas-phase nanoparticle synthesis techniques have an advantage since particles can be grown pristine while capping agents can be introduced later at will. Continuous aerosol synthesis has other advantages too, notably, the ease of altering surface termination, creating core-shell structures, and depositing nanoparticles directly onto a substrate. 6 Like most particle synthesis techniques, however, aerosol synthesis often results in broad particle size distributions that require further size-selection to ensure monodispersity. 7 Limitations in the resolution of state-ofthe-art aerosol classifiers hinder further the quest for nanoparticles less than 2.5 nm in size. Such small clusters made of silicon are particularly desirable for the study of quantum confinement effects on radiative electron-hole pair recombination and for exploring the possibility of light emission across the visible spectrum based on Si. There are, however, only few reports of blue-light emitting Si nanoparticles. 7,8 In this paper, we report on a simple, inexpensive, continuous flow, gas-phase synthesis technique, which can be used * Corresponding author. E-mail: giapis@cheme.caltech.edu. for the generic (i.e., material independent) production of nanoparticles less than 2.5 nm in size. The technique is based on high-pressure microdischarges, operating as very short residence time (µs-ms) microreactors. Microdischarges are boundary-dominated plasmas, scaled down to below 200 µm in at least one dimension, which are characterized by large electric fields. Small powers (5-10 W) are dissipated in plasma volumes of typically less than 1 µl, resulting in power densities as high as 10 KW/cm 3, unprecedented in plasma processing technology. Previous work 9,10 has established that such microdischarges contain high concentrations of energetic electrons which permit rapid decomposition of gaseous precursors for the efficient production (large gradients) of radicals in a spatially limited reaction zone. Under appropriate precursor saturation conditions, particles can nucleate and grow in the discharge region. As the particles are swept by the flow outside the discharge, radicals are no longer being produced and particle growth is terminated. Furthermore, particle charging in the microdischarge can reduce particle coagulation downstream of the reaction zone. The large precursor concentration gradients, short residence time, and charging of particles in the discharge result in the formation of very small (1-3 nm) nanoparticles with narrow size distributions. In this paper, we report on the application of the microdischarge synthesis technique in the production of high-quality Si nanoparticles, as evidenced by their photoluminescence properties and quantum efficiency. The Si nanoparticles were produced as an aerosol using a direct current (dc) microdischarge reactor, drawn schematically in Figure 1. The setup operates at a pressure slightly 10.1021/nl0480060 CCC: $30.25 Published on Web 02/04/2005 2005 American Chemical Society
Figure 1. Schematic diagram of the experimental apparatus used to synthesize silicon nanoparticles. The microdischarge forms in the cathode tip and extends a short distance outside toward the anode. Particles are either directed into a differential mobility analyzer to be classified or collected by directing the aerosol stream onto a substrate or bubbling it into a liquid. larger than 1 atm and, therefore, there is no need for a vacuum pump. The microdischarge forms in a hollow cathode (stainless steel capillary tube, i.d. ) 180 µm) and extends toward an anode (metal tube, i.d. ) 1 mm). The electrodes are separated by a gap of 1.5 mm and are pressuresealed inside a quartz tube. Typical voltages and currents used to sustain the discharge were between 300 and 500 V and 3-10 ma, respectively. For the experimental results reported here, the discharge current was kept constant at 7 ma. Silane in argon was used as the precursor gas; the silane concentration was controlled between 1 and 5 ppm by varying the flow rate of a 50 ppm SiH 4 /Ar mixture (electronic grade, Matheson Tri-Gas) while maintaining a constant total flow rate with a balance of argon. Pure argon gas was also supplied in the afterglow region outside the cathode to quench particle growth and reduce particle coagulation downstream. The size distribution of the aerosol nanoparticles was measured in-situ using a radial differential mobility analyzer (RDMA). 11 This instrument detects charged particles and is typically preceded by a bipolar charger (sealed 85 Kr β-source) to ensure proper charging of the particles. We chose not to use the charger after we discovered that it enhanced particle coagulation thus shifting the distribution to larger sizes. Instead, the plasma-charged particle stream was directed straight into the RDMA, which could then measure the distribution of particles of either charge polarity. As-grown particles were collected either on a molybdenum substrate or in a liquid without size selection for further optical and spectroscopic characterization. Dispersions of particles were obtained by bubbling the aerosol stream through a glass frit into an organic solvent, which was previously outgassed for 1-2 h to remove all dissolved oxygen. We selected 1-octanol as solvent because it has been previously shown to stabilize and passivate the surface of silicon particles. 3 After collecting particles for 24 h, the solvent was removed by vacuum evaporation to concentrate the particles. The particles were then redispersed in a nonpolar organic solvent (hexane). Energy dispersive spectroscopy and micro-raman spectroscopy were used on directly deposited particles on a Mo substrate to verify that the particles consisted of Si (see Figure 2. Size distributions of positively and negatively charged particles at two different silane concentrations as classified by a radial differential mobility analyzer. Log-normal distribution fits are also shown. Plasma reactor conditions: total flow rate through the discharge ) 150 sccm, Ar quench gas flow rate ) 450 sccm, electrode gap ) 1 mm, discharge current ) 7 ma. Supporting Information). Transmission electron microscope (TEM) imaging of silicon particles deposited from the aerosol onto carbon-coated copper grids revealed particle aggregates between 5 and 10 nm in size, probably formed during the deposition process. Electron diffraction from these samples was inconclusive as to the crystalline nature of the particles (see Supporting Information). We suspect that the coagulated particles were oxidized during air transfer to the microscope reducing the core size to below the resolution limit of the TEM. Further attempts to measure particle sizes were made by using an in-line RDMA. This instrument classifies particles based on their electrical mobility in a carrier gas, which corresponds to the projected area of the particles. Representative np-si size distributions of both charge polarities are shown in Figure 2 for two silane concentrations. In the range of silane concentrations explored here, the discharge was stable and the particle size distributions were reproducible. Below a silane concentration of 1 ppm, particles could not be detected, presumably because they were smaller than the 2.5 nm detection limit of the instrument. As the silane concentration was raised from 2.5 to 4.0 ppm, the mean particle size increased and the size distribution broadened significantly (see Figure 2). Fitting the obtained distributions to a log-normal function provided estimates of the geometric mean particle diameter (D g ) and standard deviation (σ g ). At a silane concentration of 2.5 ppm, D g and σ g were found to be 2.9 nm and 1.32, respectively. The observed σ g compares favorably with values measured by other growth processes without size selection. 11-13 The higher silane concentration of 4.0 ppm resulted in distributions with larger D g and σ g of 6.2 nm and 1.45, respectively. The observed trends and the overall shape of the size distributions (log-normal fit) are consistent with particle growth by coagulation. Note that the number density of the positively charged particles detected exceeds that of the negatively charged particles (Figure 2). This observation is consistent with Si cluster nucleation in low-pressure plasma discharges 538 Nano Lett., Vol. 5, No. 3, 2005
Figure 3. (a) Atomic force microscopy (AFM) image of Si nanoparticles drop cast from a hexane dispersion onto a Si(100) polished substrate. (b) Corresponding histogram of Si nanoparticle heights measured from the AFM image in (a). where crystallites smaller than 2 nm were barely detected using laser photodetachment. 14 Remarkably, the peak number density for each charge polarity does not change appreciably with silane concentration. The latter observation suggests that the particle density exceeds that of ions and electrons in the microdischarge available for attachment and, thus, the particles must be at most singly chargedsa requirement for the correct interpretation of the RDMA results. Large particles (g10 nm) charge up negatively in plasmas as a result of the larger electron mobility. Thus, observation of both charge polarities corroborates the existence of very small particles. Since the TEM and RDMA results suggested particle growth by coagulation, size measurements by other techniques were also attempted to estimate the actual size of the nanoparticles grown in the microdischarge. Atomic force microscopy (AFM) was performed on np-si samples, capped with octanol and suspended in hexane, after drop-casting on a silicon wafer and solvent evaporation. Dilution of the particle solution prior to drop-casting permitted the observation of large areas of noncoagulated particles as shown in Figure 3a. From the corresponding AFM histogram, plotted in Figure 3b, the mean particle height was estimated to be 1.6 nm, with overall heights as small as 1.2 nm and no larger than 6 nm. We consider this result to be a more realistic representation of the size distribution of the microplasmasynthesized nanoparticles, as corroborated further by their optical properties discussed below. Photoluminescence (PL) measurements were performed at room temperature on both liquid-suspended and directly deposited np-si samples using different setups. For octanolcapped np-si suspended in hexane, excitation and emission spectra were obtained using a commercial spectrophotometer (excitation: 75 W cw Xe lamp followed by an f/4 0.2-m Czerny-Turner grating monochromator with 0.2 nm resolution; emission collected using an identical monochromator and photon counting electronics). Figure 4 illustrates representative spectra from nanoparticles grown with 2.5 ppm silane in argon. The spectra exhibit an excitation peak at 360 nm (dashed line) and a PL emission maximum at 420 nm (solid line). The blue emission was readily observable Figure 4. Room-temperature absorbance (dotted line), PL excitation (dashed line), and PL emission spectra (solid line) of silicon nanoparticles in hexane solution. The PL emission spectrum is obtained with fixed excitation wavelength at 360 nm while the excitation spectrum is collected by fixing the detection at 420 nm. by the naked eye. In the absorbance spectra, collected using a standard UV-visible absorption spectrometer and also plotted in Figure 4 (dotted line), there is a well-defined peak at approximately 320 nm (3.9 ev), which has been previously associated with the direct transition Γ 25 -Γ 2 in 1.8 nm Si nanocrystals. 2 PL spectra of directly deposited np-si thin films were acquired by exciting with a GaN laser (405 nm) and collecting the emission with a 27.5 cm focal length grating monochromator equipped with a cooled chargecoupled device detector. Figure 5 (dashed line) illustrates PL emission spectra from such a nanoparticle film, which exhibit a broad peak at 511 nm. This peak is not directly comparable with the 420 nm PL emission maximum of Figure 4 for the hexane-suspended np-si because of the difference in excitation wavelength. New PL emission spectra were collected for the suspended np-si using lamp excitation at 405 nm. The PL emission was found to peak at approximately 465 nm, i.e., at a shorter wavelength than that for the directly deposited nanoparticles. We attribute the difference to surface oxidation resulting from exposure to Nano Lett., Vol. 5, No. 3, 2005 539
Figure 5. Room-temperature PL spectra of Si nanoparticles deposited on molybdenum substrate (dashed line), excited by a 20 mw GaN laser line at 405 nm. The sharp peak at 438 nm is an artifact of the cutoff filter used in the setup. Shown for comparison is a PL spectrum of the octanol-capped Si nanoparticles in hexane solution also excited at 405 nm using a lamp. air prior to the collection of the PL spectra. Doubly bonded oxygen at the surface of small Si clusters has been predicted 15 to introduced defect states in the gap that permit radiative electron-hole recombination at energies smaller than the cluster band gap. Singly bonded oxygen termination of the Si surface, as it must be the case for the octanol-capped nanoparticles, has been predicted 15 to result in shallower defect levels, consistent with our observation of the blueshifted emission for the suspended np-si. Assuming that the PL emission at 420 nm (2.95 ev) is from band-to-band recombination, the silicon particle core size can be estimated from calculations 15-17 to be about 1.7 nm. The actual size could be even smaller, given the octanol capping of the particle surface, 15 which corroborates the AFM size estimation. The larger sizes detected with the RDMA and observed in the TEM could be attributed to particle coagulation during aerosol classification or grid deposition. We note that particles grown at higher silane concentrations, which appear to be bigger according to the RDMA (see Figure 2), do not exhibit red-shifted PL peaks as it would be expected from quantum confinement. This observation lends further credibility to the hypothesis that the primary particles formed in the microdischarge are limited in size in the 1-2 nm range, where particle coagulation is likely unless the particles are immediately capped and dispersed in a solvent. Thus, larger silane concentrations result in the production of more particles in the same size range. The radiative lifetime τ of the PL emission of as-grown np-si was also evaluated. PL decay measurements were performed at 295 K by exciting np-si in hexane with an Ar ion laser at 457.9 nm with a cw power density of 5 mw/ mm 2 at the sample. The PL emission was monitored at 570 nm through a cutoff filter at 500 nm. The beam was pulsed at 1000 Hz using an acoustic optical modulator (50% duty cycle) and the data were collected with a temporal resolution of 5 ns. The results, shown in Figure 6, could be approximated by a stretched exponential with τ ) 30 ns and a Figure 6. Photoluminescence decay of emission at 570 nm from a Si nanoparticle dispersion in hexane solution (λ ex ) 454.8 nm). The solid line represents fitting to a stretched exponential with parameters τ (radiative lifetime) and β (exponent). stretch fitting parameter β ) 0.9. The obtained lifetime of 30 ns is longer than the e1 ns lifetimes typically attributed to surface state recombination of electron-hole pairs 2,19 but shorter than transitions believed to be associated with bandto-band recombination (10-100 µs). 5,18 While the light emission mechanism in np-si is still under debate, it seems that both theory 20 and experiment 21 agree on a trend of decreasing PL lifetimes for smaller nanoparticles. Given the size of our Si nanoparticles, short lifetimes should be expected, perhaps even shorter than the 30 ns measured, which was based on an excitation wavelength that is significantly red-shifted from the excitation optimum. It has not been assessed yet whether changes in surface termination will influence PL lifetime. Reported values for the external quantum efficiency of np-si have ranged from less than 1% to as high as 23%. 2,3,5,8 To estimate the quantum efficiency of the microdischargesynthesized Si nanoparticles, PL emission spectra were collected for various particle dilutions. The integrated emission intensity for each sample is plotted in Figure 7 against the corresponding absorbance at the same excitation wavelength. Identical measurements (excitation conditions, lamp energy, and spectrometer band-pass) were performed on 9,10-diphenylanthracene in cyclohexane, which emits between 400 and 500 nm with a known efficiency of 90%. A comparison of the slopes of the linear fits to the two sets of data, shown in Figure 7, suggests that the quantum efficiency of the octanol-capped Si nanoparticles was 30%. We emphasize that these are as-grown nanoparticles without any further processing or special treatments. In summary, high-pressure microdischarges have been used as short residence time microreactors to synthesize nanometer-size silicon particles from silane. In-situ size classification of the aerosol formed in the discharge indicated strongly coagulated particles with sizes between 2 and 5 nm distributed narrowly about the mean (σ g ) 1.3). AFM characterization of solvent-collected and dispersed particles suggested formation of individual Si clusters with sizes between 1 and 2 nm. As-grown Si nanoparticles, capped with 540 Nano Lett., Vol. 5, No. 3, 2005
Supporting Information Available: TEM image of silicon nanoparticles deposited from aerosol onto a carboncoated copper grid and spectral characterization of np-si film deposited on Mo substrate. This material is available free of charge via the Internet at http://pubs.acs.org. Figure 7. Integrated PL intensity versus absorbance for multiply diluted Si nanoparticles in hexane and for various dilutions of 9,- 10-diphenylanthracene in cyclohexane under identical excitation conditions. The quantum efficiency of the Si nanoparticles is calculated to be 30% from the ratio of the slopes of the linear fits to the experimental points times the known quantum yield of diphenylanthracene (90%). octanol, exhibited room-temperature photoluminescence with a peak at 420 nm and a quantum efficiency of 30%. The microdischarge synthesis is a simple, inexpensive, benchtop technique that could be used to synthesize continuously nanoparticles of generic composition from gaseous precursors. Acknowledgment. The authors are grateful to Julie Biteen for assistance with the PL lifetime measurements and to the Tirrell group for the use of their fluorescence spectrometer. This article was based on work supported by NSF (CTS-0404353). References (1) Yoshida, T.; Yamada, Y.; Orii, T. J. Appl. Phys. 1998, 83, 5427. (2) Wilcoxon, J. P.; Samara, G. A.; Provencio, P. N. Phys. ReV. B1999, 60, 2704. (3) Holmes, J. D.; Ziegler, K. J.; Doty, C.; Pell, L. E.; Johnston, K. P.; Korgel, B. A. J. Am. Chem. Soc. 2001, 123, 3743. (4) Belomoin, G.; Therrien, J.; Smith, A.; Rao, S.; Twesten, R.; Chaieb, S.; Nayfeh, M. H.; Wagner, L.; Mitas, L. Appl. Phys. Lett. 2002, 80, 841. (5) Littau, K. A.; Szajowski, P. J.; Muller, A. J.; Kortan, A. R.; Brus, L. E. J. Phys. Chem. 1993, 97, 1224. (6) Ostraat, M. L.; DeBlauwe, J. W.; Green, M. L.; Bell, L. D.; Atwater, H. A.; Flagan, R. C. J. Electrochem. Soc. 2001, 148, G265. (7) Orii, T.; Hirasawa, M.; Seto, T. Appl. Phys. Lett. 2003, 83, 3395. (8) Li, X.; He, Y.; Talukdar, S.; Swihart, M. T. Langmuir 2003, 19, 8490. (9) Sankaran, R. M.; Giapis, K. P. J. Appl. Phys. 2002, 92, 2406. (10) Sankaran, R. M.; Giapis, K. P. J. Phys. D. 2003, 36, 2914. (11) Camata, R. P.; Atwater, H. A.; Valhala, K. J.; Flagan, R. C. Appl. Phys. Lett. 1996, 68, 3162. (12) Suzuki, N.; Makino, T.; Yamada, Y.; Yoshida, T.; Seto, T. Appl. Phys. Lett. 2001, 78, 2043. (13) Rao, N.; Micheel, B.; Hansen, D.; Fandrey, C.; Bench, M.; Girschick, S.; Heberlein, J.; McMurry, P. J. Mater. Res. 1995, 10, 2073. (14) Stoffels, E.; Stoffels, W. W.; Kroesen, G. M. W.; dehoog, F. J. J. Vac. Sci. Technol. A 1996, 14, 556. (15) Puzder, A.; Williamson, A. J.; Grossman, J. C.; Galli, G. Phys. ReV. Lett. 2002, 88, 097401-1. (16) Hill, N.; Whaley, K. B. Phys. ReV. Lett. 1995, 75, 1130. (17) Wolkin, M. V.; Jorne, J.; Fauchet, P. M.; Allan, G.; Delerue, C. Phys. ReV. Lett. 1999, 82, 197. (18) Wilson, W. L.; Szajowski, P. F.; Brus, L. E. Science 1993, 262, 1242. (19) Lu, X.; Hanrath, T.; Johnston, K. P.; Korgel, B. A. Nano. Lett. 2003, 3, 93. (20) Delerue, C.; Allan, G.; Lannoo, M. Phys. ReV. B1993, 48, 11024. (21) Garcia, C.; Garrido, B.; Pellegrino, P.; Ferre, R.; Moreno, J. A.; Morante, J. R.; Pavesi, L.; Cazzanelli, M. Appl. Phys. Lett. 2003, 82, 1595. NL0480060 Nano Lett., Vol. 5, No. 3, 2005 541