Aerosol-Phase Synthesis and Processing of Luminescent Silicon Nanocrystals

Abstract

Silicon quantum dots are attractive materials for luminescent devices and bioimaging applications. For these light-emitting applications, higher photoluminescence efficiency is desired in order to achieve better device performance. Nonthermal plasma synthesis successfully allows for the continuous production of silicon nanocrystals, but postprocessing is necessary to improve photoluminescence quantum yields so that nanocrystals can be used for luminescence applications. In this work, we demonstrate an all-aerosol-phase synthesis and processing route that integrates nonthermal plasma synthesis, plasma-assisted surface functionalization with alkene ligands, and in-flight annealing within one flow stream. Here, luminescent silicon nanocrystals are synthesized and postprocessed on a time scale of only 100 ms, which is orders of magnitude faster than previous synthesis and functionalization schemes. The as-produced silicon nanocrystals have photoluminescence quantum yields exceeding 20%, which is a 5-fold increase compared to previous silicon nanocrystals synthesized with all-aerosol-phase approaches. We attribute the enhanced photoluminescence to the reduced “dark” nanocrystal fraction due to reduction of dangling bond density and desorption of surface silyl species induced by the in-flight annealing. We also demonstrate that the ligand coverage plays a minor role for the photoluminescence properties, but that the nature of the silicon hydride surface groups is a major factor.

Graphical Abstract

INTRODUCTION

Quantum-confined nanocrystals, also known as quantum dots, are attractive materials because of their unique highly tunable optical and electronic properties.^1-3 Among the great variety of quantum dot materials, silicon nanocrystals (Si NCs) are drawing increasing attention due to their earth abundance, biocompatibility, and low toxicity.^4,5 Silicon nanocrystal based devices have been developed including luminescent devices,⁶ light-emitting devices,^7-10 and photodetectors,¹¹ and Si NCs have been used in biomedical applications.^12,13 One of the common methods of synthesizing Si NCs is decomposition of silane in a nonthermal plasma.^14,15 In a nonthermal plasma, nanoparticles can be heated by gaining energy from recombining electrons and ions, leading to the particles’ crystallization.¹⁵ Compared to colloidal methods, plasma synthesis eliminates the need for solvents and ligands. Furthermore, particles are created with bare surfaces, which enables added control over subsequent surface treatments.

Unfortunately, as produced, plasma-synthesized Si NCs have low photoluminescence quantum yield (PLQY). For applications which rely on high PLQY, it is crucial to apply surface functionalization methods to achieve higher PLQY.^16-18 Previous research has shown that PLQY of Si NCs can be enhanced via alkene/alkyne hydrosilylation,¹⁷ which involves the grafting of ligands to the hydrogen-terminated silicon surfaces through covalent Si─C bonds. Hydrosilylation can be initiated either thermally,¹⁷ photochemically,^19-22 by applying a catalyst,²³ or using bifunctional ligands.²⁴ The hydrosilylation reaction has been shown to improve the PLQYs up to 60% in freestanding Si NCs.¹⁷ However, each of these hydrosilylation reactions is conducted postsynthesis and may require, depending on the synthesis method used, a switch from a nonliquid phase synthesis to a liquid phase postprocessing.

Conducting all synthesis, functionalization, and processing steps in the gas phase is attractive due to lack of solvents and the presumed speed of an approach that does not require a transition from the gas phase to the liquid phase. An allaerosol-phase functionalization approach was proposed by Mangolini and Kortshagen in which the nanoparticles flow through a second plasma chamber after the main synthesis plasma.²⁵ Ligands were grafted onto Si NC surfaces in the second plasma, and these functionalized nanoparticles formed stable solutions in toluene. Anthony et al. simplified this scheme by combining synthesis and functionalization in two spatially separated regions of a single plasma.²⁶ The NC synthesis was achieved in a higher density upstream synthesis region and functionalizing ligands were injected further downstream into the lower density plasma “afterglow”. This approach has the advantage that nanocrystals never leave the plasma between their synthesis and functionalization, remaining in a negatively charged state that prevents nanocrystal aggregation. However, even though ligands were successfully grafted onto Si NC surfaces, the PLQY remained low at less than 5%. The authors did observe that the PLQY could be improved when the plasma-functionalized Si NCs were heated in mesitylene, but this negated the advantage of the all-aerosol-phase synthesis and functionalization scheme.

In this work, we discuss an all-aerosol-phase synthesis and functionalization route for silicon nanocrystals with >20% PLQY that is higher than previous all-aerosol-phase approaches. Building on the work of Anthony et al.,²⁶ Si NCs are synthesized in the intense upstream zone of the plasma and ligands are grafted onto the Si NC surfaces further downstream in the weak plasma “afterglow” where the density of plasma species decays through recombination. The residence time of the NCs in the synthesis and the functionalization zones is just a few milliseconds. However, different from previous work,²⁶ Si NCs are then passed through a flow through furnace, where they are heated in flight. Such in-flight heating was previously used by Alvarez Barragan et al. to grow carbon shells on silicon crystals for lithium ion batteries.²⁷ The heat capacitance of the 2–4 nm Si NCs is small, and the characteristic time for the NC temperature to equilibrate with the gas temperature through conduction is on the order of 0.01 ms, even considering the low-pressure environment at which the Knudsen effect becomes significant.¹⁵ By letting the aerosol stream of Si NCs flow through a tube furnace where the residence time is ~100 ms, the temperature of Si NCs will closely follow the gas temperature profile in the furnace. As we discuss below, this in-flight annealing significantly improves the PLQY. Hence, with this all-aerosol-phase design, highly luminescent Si NCs are synthesized on a time scale of 100 ms, orders of magnitude faster than schemes that require liquid postsynthesis processing.

EXPERIMENTAL SECTION

Si NC Synthesis and Functionalization.

A schematic of the experimental setup for this study is shown in Figure 1. Silicon nanoparticles are synthesized and functionalized in a nonthermal radio frequency (rf) plasma reactor as described in ref 26. Argon and silane (5% in helium) enter through a borosilicate glass tube reactor with ligands introduced into the plasma afterglow by flowing hydrogen through a bubbler to carry the ligands. Hydrogen is chosen as the carrier gas since previous studies showed that Si NCs have lowest defect densities and best photoluminescence quantum yields when hydrogen is injected into the plasma afterglow compared to argon and helium.²⁸ Radio frequency power at 13.56 MHz, and a nominal value of 60 W was applied to a pair of copper ring electrodes placed 2 cm before the afterglow region. The argon flow rate varies between 30 and 100 sccm (Table 1), while silane (5% in helium) and hydrogen flow rate were fixed at 14 and 100 sccm, respectively. Ligands used in this study included 1-octene, 1-dodecene, and 1-octadecene (from Alfa Aesar, TCI, and Sigma-Aldrich, respectively). The gas pressure in the plasma region was 2.2–2.4 Torr, and gas pressure in the bubbler was 35–100 Torr. After the plasma region, the gas stream was sent through a tube furnace, with temperature profiles shown in Figure 2. Si NCs were collected onto glass substrates by inertial impaction using a setup similar to that used in ref 29. The typical collection rate of 3.2 nm, 1-octene-functionalized Si NCs was around 10 mg/h. Additional details are provided in the Supporting Information.

Figure 1.

Schematic of the experimental setup in this study. The three stages of nanocrystals growth and surface modification are the growth of silicon nanocrystals through dissociation of the SiH₄ precursor and particle nucleation, grafting of organic ligands onto the silicon nanocrystal surfaces, and the desorption of surface silyl species through in-flight annealing.

Table 1.

Synthesis Conditions for Si NCs in This Study^a

size (nm) (TEM)	Ar flow rate (sccm)	pressure (Torr)
2.5	80	2.4
3.2	30	2.4
4.5	20	2.2

^aFor all samples, SiH₄ (5% in He) and H₂ flow rates are 14 and 100 sccm, respectively.

Figure 2.

Temperature profiles of the tube furnace. Measurements were taken with a sheathed k-type thermocouple. To replicate typical operating conditions, 30 sccm Ar, 100 sccm SiH₄, and 14 sccm He were flown through the furnace.

Characterization.

Photoluminescence spectra were recorded with an Ocean Optics USB 2000 spectrometer in an integrating sphere setup as described in ref 30. The excitation source is a 395 nm LED, and absolute quantum yields were calculated quantitatively using the absorption and emission intensities. The spectrometer was calibrated with an Ocean Optics LS-1-CAL tungsten halogen light source. Immediately after synthesis, Si NCs were transferred air-free into a nitrogen-filled glovebox via a pushrod assembly, where they were dispersed in 1 mL of toluene at the concentration of ~1 mg/mL and sealed in a borosilicate glass vial for PLQY measurements. The PLQY measurements were then immediately performed, while the Si NCs were kept in the nitrogen-filled glass vials during the measurements. The spectrum of the glass vial containing 1 mL of toluene under the illumination of the excitation source is first recorded as the baseline. The spectrum is divided into the excitation region and the luminescence region. Once the sample is inserted, the signal from the excitation region will decrease and the signal from the luminescence region will appear. The PLQY is calculated by $\text{PLQY}=\frac{P_{\mathrm{s}}-P_{\mathrm{b}}}{-(L_{\mathrm{s}}-L_{\mathrm{b}})}$, where L_s,b = ∫_excitationλI(λ) dλ and P_s,b = ∫_luminescenceλI(λ) dλ are integrals of the spectral regions of excitation and emission, respectively, with the subscript s indicating the sample and b the baseline measurements. Unless specified otherwise, the spectra appearing in figures are normalized to the integral −(L_s – L_b) which represents how much of the excitation signal is absorbed by the sample. For Si NCs having the same size, and thus having identical absorption properties, this normalization guarantees that the luminescence spectra represent the emission from the same amount of Si NCs.

For structural and surface characterization, X-ray diffraction (XRD) data were recorded with a Bruker D8 Discover diffractometer with a Co X-ray source. Scherrer fitting was applied to determine the nanocrystal sizes using the MDI JADE software with shape factor K = 0.89. Bright-field transmission electron microscopy (TEM) images were taken using a Tecnai T12 microscope with an accelerating voltage of 120 kV. For TEM measurements, Si NCs dispersed in toluene were drop-cast onto a carbon-coated TEM grid, and measurements were taken after the solvent evaporated. Fourier transform infrared (FTIR) spectra were recorded with a Bruker ALPHA FTIR spectrometer using the diffuse reflectance (DRIFTS) mode in a nitrogen-filled glovebox. FTIR samples were deposited on aluminum-coated silicon substrates with an impactor described in ref 29. Twenty scans were taken for each measurement at 2 cm⁻¹ resolution. Electron paramagnetic resonance (EPR) measurements were performed at ambient temperature in a Bruker continuous wave EleXsys E500 EPR spectrometer equipped with X-band (9 GHz) microwave bridge and spherical SHQ resonator. An amount of 5 mg of Si NCs was dispersed in 0.1 mL of toluene for all samples. EPR grade quartz tubes with 5 mm diameter were used for the measurements, and the tubes were sealed in nitrogen to avoid air exposure during measurements. The surface ligand coverages were measured with a 2400 Series II CHNS/O elemental analyzer performed by Atlantic Microlab, Inc. Weight percentages of carbon and hydrogen are within ±0.3% error limits. The surface coverage values are estimated by calculating the ratio of the number of ligands per NC to the number of surface Si atoms per NC.

To measure the size of silicon nanocrystals suspended in solution, dynamic light scattering (DLS) measurements were taken with a Nanotrac Flex in-situ analyzer using volume distribution mode. The equipment was calibrated with NIST-traceable particle size standards, and the mean diameter measured has accuracy of within 5%. Solvents (toluene, from Fisher Scientific) used for characterization were dried over 4 Å molecular sieves and degassed by bubbling nitrogen through the solvent.

RESULTS AND DISCUSSION

Si NC Characterization.

Si NCs with sizes ranging from 2.2 to 4.5 nm are prepared using the all-aerosol-phase setup, schematically shown in Figure 1, which is described in detail in the Experimental Section. The argon carrier gas flow rate, which affects the residence time of NCs in the plasma region, is adjusted to tune the Si NC size. According to TEM images (Figure 3a), average sizes of 1-dodecene-functionalized Si NC samples are 2.5, 3.2, and 4.5 nm, not including the ligand shell. This is consistent with crystallite sizes of these Si NCs of 2.2, 3.1, and 3.9 nm based on the XRD pattern and applying Scherrer’s equation (Figure 3b). Different from Si NCs that were not functionalized and heated in the gas phase, the Si NCs produced here exhibit photoluminescence in the near-infrared (NIR) immediately after synthesis and gas-phase processing (Figure 3c). The emission peak wavelengths of 2.5, 3.2, and 4.5 nm Si NCs are around 750, 820, and 920 nm, respectively, which is in agreement with emission energies predicted by quantum confinement models.^31,32

Figure 3.

(a) TEM images of Si NCs of three sizes prepared by dispersing the as-produced samples in toluene and drop-casting on a TEM grid. The ligand is 1-dodecene, and furnace processing temperature is 500 °C. (b) XRD pattern of the 1-dodecene-functionalized Si NCs. (c) Photoluminescence spectra of 2.5, 3.2, and 4.5 nm Si NCs that are gas-phase functionalized and in-flight heated at 500 °C and dispersed in toluene. Spectra are normalized to 1. (d) Cumulative number-weighted distribution of as-produced, 1-dodecene-functionalized 3.2 nm Si NCs with 500 °C furnace temperature dispersed in toluene determined by DLS. Sample concentration is ~1 mg/mL. The inset is a photo of Si NCs dispersed in toluene.

The as-produced Si NCs form stable solutions in common nonpolar solvents including toluene and chloroform for furnace temperatures up to 500 °C. No sedimentation is observed over a period of a few months. Dynamic light scattering was performed to quantitatively examine the dispersity of Si NCs. Figure 3d shows the cumulative size distribution from DLS measurements taken for 3.2 nm Si NCs dispersed in toluene processed at the furnace temperature of 500 °C. The median diameter of the number-weighted distribution is 6.8 nm. Considering that Si NCs are capped with dodecyl, which results in a ligand shell of ~1.6 nm thickness, this distribution of hydrodynamic diameters is consistent with singly dispersed Si NCs. When the furnace temperature is raised to higher than 500 °C, a stable solution can no longer be formed, presumably because the high temperature may lead to degradation of the ligands that are required for colloidal stability.

Influence of Annealing Temperature on PLQY.

A significant enhancement in PLQY is observed when in-flight annealing is applied, with the furnace temperature profile shown in Figure 2. Considering the parabolic temperature profile and assuming 18 cm as the effective heating length, the in-flight annealing time is ~100 ms for the flow rates used here. As is shown in Figure 4, for 3.2 nm Si NCs functionalized with 1-octene, 1-dodecene, or 1-octadecene, the PLQYs increase with increasing temperature up to 500 °C, reaching a maximum of around 20% (Si NCs of other sizes display similar trends). Note here that the PLQY values need to be taken with caution, due to the low sensitivity of the spectrometer in the NIR region and possible errors from the integrating sphere method.³³ The PLQY and peak emission wavelength are not affected by ligand length (Figure S1). Similar to the result obtained by Mangolini and Kortshagen,²⁵ the PLQY of as-produced Si NCs that are not heated in flight is typically lower than 5%. For all the furnace annealing temperatures shown in Figure 4, the emission peaks are located around 828 nm, with some minor sample-to-sample variances. It is believed that red to near-infrared photoluminescence of Si NCs originates from the silicon core states and is subject to quantum confinement effects.^34,35 The fact that the peak wavelength of the photoluminescence remains constant implies that there is no shrinking of the Si NC core size associated with annealing. Further, size estimations from XRD measurements indicate that crystallite size remains constant during the heat treatment (Figure S2). This indicates that the main role of heat treatment is associated with the modification of nanocrystal surfaces.

Figure 4.

(a) PL spectra (excitation λ = 395 nm) of 3.2 nm, 1-octene-functionalized Si NC processed at different furnace temperatures. Spectra are taken with an integrating sphere, and PL intensities are normalized by dividing the spectrum by the absorption integral. Measurements were taken for Si NCs dispersed in toluene with concentration of ~1 mg/mL. (b) PLQY as a function of furnace annealing temperature for samples functionalized with 1-octene (8C-Si), 1-dodecene (12C-Si), and 1-octadecene (18C-Si).

In order to characterize changes of Si NC surfaces upon in-flight annealing, we performed EPR measurements and FTIR spectroscopy. EPR measurements are capable of identifying paramagnetic defects, including dangling bond centers, which represent an important type of trap state in silicon.³⁶ They are believed to be nonradiative centers which contribute to a reduction of the PLQY of silicon nanocrystals.³⁷ EPR measurements on 1-dodecene-functionalized Si NCs with various furnace heating temperatures were performed while carefully keeping the samples air-free and keeping sample concentration and volume constant for all samples. The EPR signal centers at 3335 G, corresponding to a g-factor of 2.007, which agrees with previous values of silicon nanocrystals functionalized in the gas phase.²⁵ The broad EPR peak has been previously attributed to a combination of g_∥ = 2.0019 and g_⊥ = 2.0086 from P_b centers at the silicon/silicon oxide interface and g_D = 2.0053 from dangling bond D centers in a disordered environment^38,39 for silicon nanocrystals showing some amount of surface oxidation. As is shown in Figure 5a, the amplitude of the EPR signal is reduced significantly as Si NCs go through in-flight annealing with furnace temperature at 300 °C. In-flight annealing at higher temperatures leads to a further reduction in EPR signal. We note that, even for the sample with the 500 °C treatment, there are still EPR-active defects that the heat treatment does not remove. While our annealing time is limited to ~0.1 s by the experimental setup, we believe that a longer annealing time in the gas phase might enable more efficient elimination of defects and thus higher PLQY.

Figure 5.

(a) EPR spectra of 3.2 nm, 1-dodecene-functionalized Si NCs with various furnace operating temperatures. The EPR signal amplitude decreases with increasing furnace temperature for the temperatures studied. (b) FTIR spectra of 3.2 nm, 1-octene-functionalized Si NCs with furnace temperature ranging from room temperature to 550 °C. (c) Histogram representing the relative area ratio of CH_x/SiH_x obtained from FTIR in panel b. (d) FTIR spectral region between 2200 and 2000 cm⁻¹ related to Si–H stretch normalized to the signal intensity at 2110 cm⁻¹. (e) Percentage of each SiH_x species obtained by peak deconvolution of the SiH_x stretching region between 2200 and 2000 cm⁻¹.

FTIR measurements provide valuable information about surface group composition. Figure 5b shows the FTIR spectra of 1-octene-functionalized Si NCs. The peaks between 2800 and 3000 cm⁻¹ are assigned to CH_x (x = 1–3) stretching modes, and the feature between 2000 and 2200 cm⁻¹ (x = 1–3) corresponds to SiH_x stretching modes of Si surface atoms that are not functionalized with a ligand. Therefore, the relative area ratio of the CH_x stretching region to the SiH_x stretching region (2000–2200 cm⁻¹) can qualitatively describe the surface ligand coverage (Figure 5c). For a set of different furnace temperatures, including room temperature, the CH_x/SiH_x area ratio remains relatively constant, which indicates minor differences in surface ligand coverage. This result indicates that, although unreacted ligands can enter the furnace along with Si NCs, further ligand functionalization does not occur in the furnace. Ligand functionalization is essentially complete once the Si NCs leave the afterglow-plasma functionalization zone.

As the increase of PLQY with increased annealing temperature is presumably not related to the attachment of additional ligands, it is reasonable to investigate the changes of SiH_x surface species caused by the annealing. Previous studies had suggested that thermal desorption of hydrogen^40,41 and changes of SiH_x species composition^42-45 can occur at the temperatures used here for in-flight annealing of the Si NCs. Most relevant for this study, Shu et al. proposed that the desorption of hypervalent SiH₃ is associated with an improvement in PLQY of plasma produced Si NCs.⁴⁶

Figure 5d shows the detail FTIR spectra of the SiH_x stretching mode region. The dominant peak in the 2000–2200 cm⁻¹ range of the sample produced without postsynthesis annealing (20 °C) is centered at ~2141 cm⁻¹, which is attributed to the SiH₃ stretch vibrations.^40,43 A reduction in the SiH₃ feature is observed at higher furnace temperatures, while the feature corresponding to SiH₂ at ~2095 cm⁻¹ becomes more significant. The drop of SiH₃ groups occurs most drastically at around 400–500 °C (Figure 5e), which is consistent with the desorption temperatures observed by Holm and Roberts for hydrogen-terminated, plasma-synthesized Si NCs.⁴⁷

The surfaces of Si NCs synthesized from a nonthermal plasma consist of silicon monohydride (SiH), dihydride (SiH₂), and trihydride (SiH₃) groups. In a previous study, it was observed that PLQY of plasma-synthesized Si NCs can be improved by heating them to 160 °C in mesitylene for an hour.⁴⁶ The proposed mechanism is that some of the SiH₃ species, produced as silane fragments in the plasma and not fully consumed during the Si NC growth, are hypervalently bonded to fully coordinated silicon atoms at the Si NC surface, and that the removal of these weakly bonded species by heating leads to a subsequent increase in PLQY. These conclusions were supported by residual gas analysis showing a low desorption temperature of SiH₃ and quantum chemical calculations showing that such hypervalently bonded species can act as nonradiative recombination centers. This mechanism is different from the desorption of a hydrogen from a covalently bonded surface SiH₃ group, which may cause dangling bond formation. Following this previous work, we here also interpret the increase in PLQY and the decrease of SiH₃ in terms of desorption of hypervalently bonded SiH₃. Compared with the previous solution phase study, the desorption process here occurs at much shorter time scales due to the much higher temperatures enabled by the in-flight heating.

Studying the temporal dynamics of photoluminescence decay provides additional information regarding the nature of the improved PLQY. Figure 6 shows the temporal photoluminescence decay of samples emitting at 820 nm, with samples emitting at other peak wavelengths behaving similarly. The time-resolved photoluminescence decay is characterized by a stretched-exponential decay I = I₀ exp(−(t/τ)^β), with lifetime τ of 100–120 μs and dispersion factor β around 0.7–0.8 (Figure 6). This slow radiative decay is characteristic of emission that originates from silicon core states following quantum confinement.⁴⁸ Interestingly, both τ and β change very little with postsynthesis annealing in the furnace, yet the PLQY increases from below 5% to around 20%. This result is consistent with previous reports by Sangghaleh et al. that an ensemble of Si NCs is composed of “bright” and “dark” quantum dots.⁴⁹ As discussed in that paper, but also pointed out in earlier studies,^50,51 the ensemble PLQY is often lower than the internal quantum efficiency IQE = Γ_r/(Γ_r + Γ_nr) of individual Si NCs, where Γ_r and Γ_nr represent the radiative decay rate and nonradiative decay rate of an individual nanocrystal. While the IQE of “bright” dots can approach 100%,^49,51,52 the ensemble PLQY also accounts for the presence of “dark” Si NCs, which have a very large nonradiative rate, i.e., which absorb but do not emit and have an IQE close to zero. The increase in ensemble PLQY with in-flight annealing at essentially constant τ and β suggests that heating turns some “dark” dots into “bright” dots, while not affecting the intrinsic radiative decay dynamics of the “bright” dots.

Figure 6.

(a) Time-resolved decay data for 1-dodecene-functionalized, 3.2 nm Si NCs with various furnace temperatures detected at 800 nm. Excitation wavelength is 407 nm, and excitation time is ~50 μs. Si NCs are dispersed in toluene with concentration of ~1 mg/mL. (b) Characteristic times τ (squares) and dispersion factors β (triangles) as a function of furnace processing temperatures.

Influence of Surface Ligand Coverage.

Here we investigate the influence of surface ligand coverage on sample PLQY with 1-octene-functionalized Si NCs. Surface ligand coverage is conveniently tuned by controlling ligand injection rate, and values of surface ligand coverage are quantified according to the weight percentage of carbon and hydrogen measured by CHNS/O elemental analysis. By assuming that ligands are covalently bonded to the silicon NC surfaces,²⁶ and that ligands remain complete without breaking into fragments or oligomerizing, surface coverage can be estimated using particle size and weight percentage of ligands. Three conditions used for 1-octene functionalization, and the corresponding surface ligand coverages are listed in Table 2. As expected, surface coverage increases with higher ligand injection rate, which is also qualitatively observed in FTIR spectra (Figure S3). While ligand surface coverages range from 25% to 43% for these conditions, the sample PLQY values remain relatively constant when transferred and measured in toluene. This observation clarifies the impact of ligand functionalization of Si NCs on the PLQY. While it is commonly believed that ligand functionalization terminates surface dangling bonds, and therefore a higher surface ligand coverage should enable higher PLQY, this is not supported by our observations. For our Si NCs produced in a nonthermal plasma, PLQY is primarily determined by surface silyl composition rather than surface ligand coverage, because only an estimated 25–43% of surface Si─H sites are replaced by Si─C bonds, due to steric hindrance of the ligands. Hence, for our plasma-synthesized Si NCs, PLQY is closely related to the postprocessing temperature which strongly affects surface SiH_x composition.

Table 2.

Three Conditions Used to Study the Relationship between Sample PLQY and Surface Ligand Coverage^a

bubbler pressure (Torr)	ligand flow rate (sccm)b	ligand mass fraction (%)	surface coverage (%)	sample PLQY (%)
100	17	23.3	25	17.0
50	35	32.0	39	16.4
35	50	35.3	43	19.0

^aFurnace processing temperature is 450 °C for all samples, and PLQY measurements are taken for Si NCs in toluene at ~1 mg/mL.

^bVapor pressure of 1-octene is 17.4 Torr at room temperature (20 °C).

However, further reduction of surface ligand coverage will have a negative influence on sample PLQY. We also measure the PLQY value and emission spectrum (Figure S4) for nonfunctionalized Si NCs, which may be of interest for device applications that require electrical conduction⁵³ due to the lack of insulating ligands. At the same 450 °C furnace processing temperature, PLQYs of H─Si NCs are only around 7%. This is likely the result of energy transfer from “bright” to “dark” nanocrystals, as we discussed above that the intrinsic sample PLQY is not primarily determined by ligand coverage. Different from the ligand-functionalized samples, the H─Si NCs remain agglomerated in toluene, resulting in much more efficient energy transfer. The interpretation is consistent with the significant red shift of the PL spectrum of H─Si NCs compared to the ligand-functionalized samples, consistent with energy transfer from smaller to larger NCs (Figure S4).

CONCLUSIONS

Si NCs with ~20% PLQY and very good solubility in nonpolar solvents have been synthesized via a one-step, all-gas-phase approach. The nanoparticles are synthesized in a nonthermal plasma and functionalized in the plasma afterglow, followed by an in-flight heat treatment in a furnace. The plasma grafting produces a ligand-functionalized surface, which grants the Si NCs solubility in common nonpolar solvents. While the heat treatment does not affect surface ligand coverage, it has a strong influence on the silicon–hydride groups at the Si NC surface. A significant increase in the ensemble photoluminescence quantum yield is observed, which is presumably associated with the desorption of hypervalently bonded SiH₃ groups at the surface.

The fact that Si NCs with respectable, though not recordbreaking, PLQY can be synthesized, functionalized, and heat-treated entirely in the gas phase on time scales of ~100 ms represents an advance that may enable the practical application of Si NCs for many luminescence applications. A scale-up of this technology is also possible through scaling reactor dimensions and/or parallelizing systems. While equipment limitations did not allow us to study longer annealing times in the gas phase, it is possible that these may enable even higher PLQY.