Vapor Phase Methods

Silicon nanoparticles can be prepared in the vapor phase by decomposition of silicon-containing gases or vapors such as silane, disilane, and chlorosilanes, or by thermal, plasma, or laser-driven vaporization (evaporation) of solid silicon followed by nucleation of silicon particles and rapid quenching to prevent their growth and agglomeration. Unwanted silicon particle formation is a common by-product of silicon film deposition by thermal or plasma chemical vapor deposition in the microelectronics industry. Studies of silicon particle formation in the gas phase in that context extend back at least to the 1971 study by Eversteijn [52] that identified the critical temperature and concentration for the onset of particle formation during epitaxial growth of silicon films from silane. There have been many studies of the formation of silicon nanoparticles in the vapor phase since then. Here, we will mainly consider those that resulted in photoluminescent particles and not the many studies of formation of larger silicon nanoparticles. The earliest reports of PL from silicon nanocrystals prepared in the vapor phase coincided with or slightly preceded the discovery of PL from porous silicon. In 1990, Takagi et al. [53] reported production of photoluminescent silicon nanoparticles a few nanometers in diameter by microwave plasma decomposition of silane. These were highly crystalline. The as-produced particles were not photoluminescent, but after heating in humid air to form a passivating oxide layer on their surface, they exhibited room temperature PL at wavelengths of 600 to 900 nm.

Brus and coworkers produced photoluminescent silicon nanocrystals by thermal decomposition of disilane in a small quartz tube in a pyrolysis oven with short residence time and collected the particles as colloids in ethylene glycol [9,10]. After surface oxidation, these nanocrystals showed size-dependent PL with broad peaks near 660, 770, and 970 nm for particles of different average size. They were able to narrow the PL emission spectra by size-separation of the nanocrystals using size-selective precipitation and size-exclusion chromatography. Their studies on these particles led to major advances in our understanding of PL from silicon nanocrystals. Unfortunately, this is probably not a practical method of making materials for further application, as their reactor produced just a few milligrams of nano-particles per 24 h day of operation. Coffer's group has used a similar pyrolysis reactor system to prepare erbium-doped silicon nanocrystals that have infrared PL emission near 1550 nm, from the erbium-dopant atoms [54,55]. Flagan's group has developed a more sophisticated version of this type of pyrolysis reactor to produce high-quality oxide-capped silicon nanocrystals for use in memory devices [56,57], but in very small quantities, as just a single monolayer of particles is needed in such devices.

From the above studies, it is clear that high-quality nanocrystals can be produced by vapor-phase decomposition of silane or disilane. However, in a conventional heated tube reactor with a residence time of order 1 s, the particle concentration must be below about 1012 particles per liter of gas to avoid collision and coagulation of the particles. A 2 nm diameter silicon nanocrystal has a mass of about 10~2° g, so this corresponds to a mass concentration of about 10 ng of silicon particles per liter of gas passing through the reactor. The coagulation rate is only weakly dependent on temperature and pressure, so at elevated temperature and reduced pressure, the volume of gas can be increased without increasing the mass flow rate. However, if the pressure is reduced too low, particle losses to the reactor wall will become unacceptably high. So, perhaps 100 ng of silicon particles per standard liter of gas (volume of 1 L at 273 K and 1 bar) can be produced. Thus, about 10 million standard liters of carrier gas would be required to produce 1 g of 2 nm diameter silicon nanoparticles. This corresponds to about 1600 size "A" cylinders of helium. This decreases with the cube of the particle diameter, so just 200 size "A" cylinders would be needed per gram of 4 nm particles. This is consistent with the Brus' group observation that their reactor system, with a somewhat shorter residence time, consumed about 1 size "A" tank of helium per 24 h of operation to produce a few milligrams of silicon nanocrystals [9]. Thus, it is clear that to produce practically useful amounts of silicon nanoparticles in the vapor phase in a reasonable time with reasonable gas consumption and reactor volume, something must be done to either reduce the residence time (by rapidly cooling the gas immediately after particle nucleation) or retard coagulation of the particles. Shorter residence times, or rapid quenching, can be achieved by laser-or plasma-induced heating. Plasma processes can also slow coagulation by producing particles that are all negatively charged and thus repel one another.

One laser-based approach is laser ablation or laser vaporization of solid silicon into a background gas. El-Shall and coworkers have developed a laser vaporization-controlled condensation (LVCC) method in which they focus a frequency-doubled Nd:YAG laser on a silicon target to generate silicon vapor in a background gas [58-61]. The silicon vapor cools very rapidly, nucleating silicon nanocrystals that are collected in the form of web-like agglomerates on a surface above the silicon source. These particles exhibit relatively strong red PL, along with weaker blue PL. Their red PL blueshifted with increased oxidation, consistent with the quantum-confined size-dependent luminescence seen in silicon nanopar-ticles prepared by other methods. Several other groups have also used laser ablation methods to produce silicon nanocrystals. Among those, the work by Makimura et al. [62] is particularly notable because they observed PL from the nanocrystals within the laser ablation chamber. When H2 was included in the background gas, they apparently formed hydrogenated silicon nanocrystals that had green PL.

Rather than using laser energy to locally heat and vaporize a solid target, one can use laser energy to dissociate vapor-phase precursor molecules. As early as 1982, Cannon et al. [63,64] used a CO2 laser to heat silane-containing gases and produce silicon nanoparticles, though these were too much large to exhibit PL. Fojtik et al. [65] used a focused ruby laser to induce breakdown and generate a small volume of plasma in an argon-silane mixture and thereby form silicon nanocrystals. These particles did not initially luminesce, but exhibited red luminescence after etching them with HF and then exposing them to air. Some blue luminescence was also observed. Huisken and coworkers have used a pulsed CO2 laser to heat silane-containing gas mixtures and generate photoluminescent silicon nanocrystals [66-73]. This produces the nanoparticles as a cluster beam that can be size-separated based on cluster velocity (the smaller particles moving faster than larger ones). These particles do not luminesce immediately after synthesis, but after exposure to air for some time, they exhibit size-dependent PL ranging from orange-yellow to the near IR. The PL continues to blueshift with continued exposure to air, as the surface oxidizes and the crystalline silicon core shrinks. The quantum efficiency of these samples appears to be quite high—30% at minimum, and perhaps much higher [73]. However, this method based on pulsed heating has the obvious disadvantage of a low duty cycle. The pulsed CO2 laser emits a less than 1 ms pulse, which allows for very rapid heating and cooling of the gases, but when operated at typical repetition rates of a few hertz, results in the formation of very small amounts of material. Botti and coworkers used a continuous CO2 laser to decompose silane and produce silicon nanocrystals [74-78], similar to the earlier work by Cannon et al. [63,64]. By diluting the silane precursor, reducing the total pressure, and other variation of reactor conditions, they were able to reduce the nanoparticle size to the point where some PL was observed.

The final vapor-phase method to be considered here is plasma-based synthesis. This is particularly promising, not only because it can be used to achieve short residence time and selective heating of precursor gases without heating reactor walls, but also because it should initially produce negatively charged silicon nanoparticles that will repel each other and thereby reduce coagulation rates. Two recent successes in this area have been presented by Mangolini et al. [79] and Sankaran et al. [80]. Sankaran et al. [80] used an atmospheric-pressure microdischarge with dimensions of order 1 mm3 (1 mL) to dissociate a mixture containing 1 to 5 ppm silane in argon. This produced nanocrystals with relatively narrow size distributions and mean diameters as small as 1 to 2 nm. These exhibited blue PL with a reported quantum efficiency of 30%. Although a single microdischarge can produce only very small quantities of nanoparticles (tens of micrograms per hour), one can imagine constructing massively parallel arrays of these discharges that could produce macroscopic quantities. Mangolini et al. [79] used a reduced-pressure nonthermal plasma in which the silane-argon gas mixture has a residence time of a few milliseconds. Particles deposited on the quartz tube downstream of this plasma discharge exhibit red to near-infrared PL after a few minutes of exposure to air. The particles had mean diameters of 3 to 6 nm, controllable by the silane partial pressure and the residence time in the discharge. Most importantly, this method produced luminescent particles at a rate of tens of milligrams per hour, among the highest production rates of all known methods for producing luminescent silicon nanocrystals.

0 0

Post a comment