High-yield green fabrication of colloidal silicon quantum dots by low-temperature thermal cracking of porous silicon

Crystalline semiconductor nanoparticles with a diameter range less than that of an exciton Bohr radius called semiconductor quantum dots (QDs) are known to have specific physical properties of a confined electron system. Size dependent bandgap and fast carrier recombination have led to a range of potential applications to optoelectronics.^1,2 In the case of silicon (Si) QDs, the breaking indirect bandgap nature allows us to exhibit a size tunable and efficient visible luminescence at room temperature.^3–5 Si QDs, which meet environmental and resource requirements, are absolutely useful materials for light-emitting devices^6–8 and biomedical applications.⁹ In particular, a colloidal form of Si QDs, i.e., those dispersed in solvents, has recently attracted much attention owing to the compatibility with the printing process, which is rapidly developing for flexible electronic devices.¹⁰ 

Colloidal Si QDs have been prepared via various bottom-up and top-down processes. Traditional bottom up processes are wet-chemical synthesis,¹¹ plasma synthesis,¹² and laser pyrolysis.¹³ In these processes, the Si precursor is directly transformed into the colloidal dots during crystal growth in each environment. A modified bottom up process has been reported where Si QDs are indirectly formed, i.e., by high-temperature heat treatment of a silicon rich SiO₂ matrix to produce Si QDs and a subsequent chemical etching process for removing the matrix.^14–17 The merit of bottom-up processes is the relatively high preparation yield per experimental batch, up to the production of Si QDs in the gram scale.

In top-down processes, the fragmentation of the bulk Si crystal can directly yield the colloidal Si QDs. Various fragmentation techniques were employed such as mechanical ball milling,¹⁸ chemical etching,¹⁹ and pulsed laser irradiation in liquid.^20–22 Besides bulk Si, an interconnected wire-nanostructure Si assembly, i.e., porous silicon (PSi),³ was used for a fragmentation target. In this case, Si QDs were indirectly formed from bulk Si through the chemical etching of bulk Si and subsequent fragmentation treatments applied to PSi. Because PSi is a relatively fragile material due to the existence of nanopores,²³ PSi is transformed into colloidal Si QDs with the use of some mild fragmentation treatments such as ultrasonification,^24,25 mechanical milling,²⁶ and unfocused pulsed laser irradiation.²⁷ In these top-down processes, the crystalline quality of Si QDs can be improved by using integrated circuit grade bulk Si crystals as a starting material. However, a scalable formation of Si QDs using a top-down process is more difficult than the bottom-up approaches.

In addition to the formation yield and the power effectiveness, a common subject remains in the above-mentioned conventional processes regarding the surface passivation control, which closely relates to the dispersibility. The stable dispersion of non-aggregating Si QD colloids should be sought in either polar or nonpolar solvent without affecting the optical and chemical activities. The key issue in this is to manage the surface bonding of Si QDs through some self-regulating processes.

With the aim of providing a practical pathway for overcoming these subjects, we present here a simple top-down process with a high transformation efficiency of colloidal Si QDs from the PSi target. Previously,²⁸ we showed that the pulsed laser irradiation to Si powder treated with stain chemical etching yields colloidal Si QDs. The formation mechanism in this process is laser-induced heating and resultant fragmentation of the etched nanostructures with significantly lowered thermal conductivity and volumetric heat capacity.²⁹ The original stain-etched Si powder, however, has a mixed structure of nanoscale Si and bulk crystalline Si.³⁰ Because the bulk Si is hardly fragmented, a large amount of waste of the raw Si powder remains after the laser irradiation process. To implement the scalable production of Si QDs for green nano-engineering, it is important to reduce waste thoroughly in the top-down process. In this work, we employ PSi flakes obtained from the electrochemical etching of the bulk crystalline Si substrate followed by the peeling off process. The mechanically fragile nature of PSi²³ should make it possible to transform into Si QDs under power- and cost-effective treatment than the usual laser induced ablation. Accordingly, simple thermal transformation is applied to PSi flakes in an organic solution with a reactive solvent (hydrofluoric acid: HF) at ∼50 °C to confirm the correlation between in situ surface passivation of Si QDs and their dispersibility in solvents.

The present process is illustrated in Fig. 1. PSi layers with a porosity of ∼80% were prepared by a standard electrochemical anodization of bulk silicon wafer in ethanol/HF mixed aqueous solution with the volume concentration of HF (50%):ethanol = 1:1. The Si substrate used was p-type Si (100) wafer, 1 Ω cm–10 Ω cm. The anodization current density and time are 50 mA/cm² and 120 min, respectively. The formed PSi layers were peeled off from the wafer by applying a high current density and drying in air. For preparing colloidal Si QDs, 5 mg–10 mg of PSi flakes were dipped in an organic solvent (3 ml) mixed with HF aqueous solution (0.3 ml). The organic solvent used was 1-decene or ethyl 10-undecenoate. The mixture content of HF in the solutions was varied in the range of 0%–50%. Then, heat treatment was done on the PSi flakes dispersed in solution for 2 h using a magnetic stirrer equipped with a hot plate. The temperatures to set the hot plate were up to 53 °C. For comparison, some samples were treated with pulsed laser irradiation at 266 nm from a Q-switched Nd:YAG laser (a pulse duration of 5 ns and a repetition rate of 15 Hz) in the same way as reported in the literature.²⁸ Subsequently, the supernatant part of the solution was extracted and, then, the unreacted residual PSi flake waste was removed from the solution by centrifugation at 16 000 rpm with 200 nm pore size membrane filters. Finally, organic solvents were removed by using a rotary evaporator or a dialysis process, and then a transparent colloidal solution dispersed in chloroform or ethanol was obtained.

Fourier transform infrared (FTIR) measurements were carried out for the QD solution samples deposited on NaCl crystal films to investigate their surface chemistry using a FTIR spectrometer (Shimadzu, FTIR-8400S). Photoluminescence (PL) spectra were measured by using a single monochromator system equipped with a charge-coupled device (Princeton Instruments, PIXIS 100B) and a He—Cd laser (Kimmon, IK3302R-E) excitation source. The PL time transient decay curves were obtained by using a frequency-tripled 355-nm light pulse from a Nd:YAG laser (Teemphotonics, STV-01E) and a Peltier-device-cooled photomultiplier tube (Hamamatsu, R375) detector. The absolute PL quantum efficiencies were measured by using a spectrofluorometer equipped with an integrated sphere (JASCO, FP-8500). These optical measurements were performed on solution samples in a quartz cuvette at room temperature.

The transmittance electron microscopy (TEM) measurements, shown in Fig. 2, reveal that the fabricated colloids are single crystalline Si nanoparticles with a diameter of 3 nm–10 nm, i.e., Si QDs. Compared to PSi flakes precipitating on the bottom of the glass tube, the prepared colloidal Si QDs exhibit efficient red PL under UV illumination, as shown in Fig. 3(a).

Figure 3(b) shows the measured PL spectra of original PSi and colloidal Si QDs prepared in 1-decene by heat treatment at ∼50 °C. Broad emission bands are observed for both samples, and the colloidal sample has the PL band at shorter wavelength side than PSi. The spectral bandwidths for both samples are almost similar: ∼300 meV. These emissions can be attributed to the band-to-band transition of electron–hole pairs confined in the nanocrystalline Si core.^31,32 Because the bandgap energy in the quantum confined system depends on its size, the difference in the PL peak energies shown in Fig. 3(b) suggests that the Si core size of the colloidal Si QDs is smaller than PSi. However, the present heat treatment at ∼50 °C should cause no uniform size reduction. More certain reason for the PL peak difference is surface-induced strain reported by K$u○$sovà et al.³³ They showed that alkyl-bonds attached on the surface of Si QDs induce tensile strains, resulting in the modification of their band structure as the energy upshift of the conduction band minimum Δ₁. Their result reasonably corresponds to the fact that the hydrostatic compressive pressure applied to most bulk indirect semiconductors including Si leads to the decrease in the bandgap energy due to the decrease in the interatomic distance, in contrast to direct semiconductor such as GaAs.³⁴ Since the surface of the present Si QDs is terminated with alkyl species and silicon QDs are assumed to inherit the indirect bandgap nature, the present energy blue-shift is presumably due to the induced tensile strains. In fact, K$u○$sovà et al. theoretically calculated the energy upshift by surface induced tensile strain to be 280 meV, which is similar to our observed PL energy shift (250 meV). The absolute PL quantum efficiency of Si QDs reaches 21%, which is seven times higher than that of PSi. This enhanced PL efficiency is attributed to a stable termination and improved crystalline feature at the surface induced by alkyl bonds, resulting in low a non-radiative recombination rate.³⁵ 

To investigate the recombination rates of the samples, we measured the dynamic PL behavior. As shown in Fig. 3(c), the PL decay of Si QDs is clearly slower than that of PSi. By fitting the PL decay curve I(t) with the stretched exponential function I(t) = I(0)exp[−(t/τ)^β], where I(0), τ, and β are initial PL intensity, PL lifetime, and stretching exponent factor, respectively,³⁶ we obtained the spectral curves of τ and β as plotted in Fig. 3(d). The τ and β values for colloidal Si QDs are larger than those of PSi in the overlapping wavelength region, indicating lower non-radiative recombination rates for Si QDs than PSi. We calculated the average decay rates w defined as the equation^37,38 w = βτ⁻¹Γ(β⁻¹), and, then, the estimated radiative rate w_r and non-radiative one w_nr from the relation Φ = w_r/w = w_r/(w_r + w_nr), where Φ is the PL quantum efficiency. The calculated values are w_r = 4.1 ms⁻¹ and 1.0 ms⁻¹ for Si QDs and PSi samples, respectively, and the corresponding w_nr = 25 ms⁻¹ and 100 ms⁻¹. The larger radiative recombination rate of Si QDs closely relates to the increase in the oscillator strength by their alkyl termination reported in the literature.³³ In addition, the lower non-radiative recombination rates indicate smaller surface defects in Si QDs.

The prepared colloidal Si QDs exhibit different dispersibilities in organic solutions depending on the preparation condition. Figures 4(a) and 4(b) demonstrate the different dispersibility of Si QD samples prepared by the heat treatment of PSi flakes in 1-decene and ethyl 10-undecenoate, respectively. The samples are well dispersed in both non-polar (hexane) and polar (methanol) solvents. Figure 4(c) shows the FTIR spectra of the Si QD samples and original PSi flakes. In the QD sample prepared in 1-decene, alkyl-ligand related signals appear at 890 cm⁻¹ (Si—C) and 1480 cm⁻¹ (Si—CH₂), although hydrogen (Si—H) and oxide (Si—O—Si) bonds originating from the PSi flakes remain. The dispersibility in the non-polar solvent is mainly due to the non-polar nature of the alkyl ligand. On the contrary, the QD sample prepared in ethyl 10-undecenoate exhibits the absorption peak at 1710 cm⁻¹, corresponding to C=O stretch vibration,³⁹ in addition to Si—C bonds at 890 cm⁻¹. This indicates that the organic ligands with the carbonyl group, i.e., the ethyl-ester group (COOR), are attached on the surface of Si QDs prepared in ethyl 10-undecenoate. The use of a polar function group makes the dispersibility of Si QDs apparent in the polar solvent shown in Fig. 4(b).

The passivation mechanism of the organic function groups is caused by hydrosilylation reaction between the organic solvents used (1-decene and ethyl 10-undecenoate) and hydrogen terminated Si surface.⁴⁰ Usually, such a hydrosilylation reaction needs a high-temperature treatment above 100 °C or UV light irradiation. However, in the case of colloidal nanocrystalline silicon, it has been demonstrated that the reaction occurs effectively even at room temperature.³⁹ Thus, our moderate-temperature heat treatment to prepare the colloidal Si QDs from PSi promotes the hydrosilylation reaction, in parallel to the morphology transformation of PSi flakes into QDs. This self-regulated concurrent effect is very useful for simplifying the fabrication process of surface controlled QDs. The conventional surface modification through the hydrosilylation reaction has been attained by separate two-step routes, i.e., the formation of hydrogen terminated Si QDs and subsequent surface modification treatments.^39,41,42

To discuss the transformation mechanism from PSi flakes to colloidal Si QDs, we evaluated the yield of PSi transformation into Si QDs, which is defined as the weight ratio of the obtained Si QDs to PSi flakes, under the varied experimental conditions. Figure 6 shows the yield as a function of the heat treatment temperature. As clearly seen in this figure, the yield of PSi transformation increases with the temperature and the yield reaches ∼125% (i.e., the amount of Si-QD colloid obtained from 5 mg to 10 mg of raw PSi is 6 mg–13 mg) at around 50 °C, indicating that the PSi flakes mostly transform into QDs. In this optimum case, the typical amount of waste consisting of residual uncracking PSi is 1 mg–2 mg. The results indicate that the Si QDs are efficiently generated by moderately heating PSi at relatively low temperatures. Note that, in the case of ultrasonification and sedimentation treatments to PSi for obtaining transparent colloidal Si-QD solution, a large part of PSi powder is filtered out as a waste.⁴³ The present process is more efficient and waste-free.

The efficiency beyond 100% means that the greater amount of QDs than the original PSi flakes, and this is due to the attachment of passivated organic molecules on the QD surface. To evaluate the amount of the attached surface ligands, we modeled single nanocrystal as a core–shell spherical nanoparticle consisting of the silicon core and organic shell. In our model, the density of each layer employs that of each bulk material, and assumed that the core diameter and shell thickness are 3.0 nm and 1.0 nm, respectively, and the coverage of the organic layer is ∼0.44, where the monolayer of alkyl groups forms on the silicon surface.⁴⁰ Based on this model, we calculated the weight ratio of the whole nanocrystals including the organic layer to the Si core to be ∼140%. This value is close to the experimental one (∼125%). The mutual difference is due to the simplified assumption in the calculation that the Si-QD surface is entirely covered with the organic monolayer. If assuming an occurrence of partial coverage, its value should then become smaller than ∼140%.

In Fig. 5, we propose a possible transformation mechanism from PSi to organically passivated Si QDs. PSi flake is an assembly of nanocrystalline Si cores. Because the PSi flake has extremely low volumetric thermal capacity and thermal conductivity,²⁹ a moderate-temperature heat treatment can induce relatively high local thermal stress in the fragile porous nanostructure network. As a result, the cracking of PSi flakes occurs and then constituent Si-QDs move apart. The bare surfaces of individual Si-QDs change into hydrides by using a HF aqueous mixture, and finally the organic function groups are passivated via hydrosilylation between the hydride Si surface and unsaturated organic molecules as mentioned above.

In our previous works,^28,44 it was reported that the pulsed UV laser irradiation to stain-etched Si powder generates the colloidal Si QDs, resulting from laser induced fragmentation of porous structures. In the case of the present PSi flakes, we confirmed that the same laser irradiation process for 2 h also promotes the formation of colloidal Si QDs, i.e., the transformation yield increases from ∼25% without laser irradiation to ∼60% with laser irradiation at room temperature (22 °C). As shown in Fig. S1 of the supplementary material, the PL spectral shape of the colloidal sample by laser irradiation is the same as that by heat treatment. Thus, the laser-induced heating of Si cores in PSi causes cracking of PSi flakes similar to the moderate heat treatment process. The lower transformation yield by the laser irradiation process than the moderate heat process is possibly due to a limited irradiation volume of the laser process. Obviously, the moderate low temperature heat process is more cost-effective than the laser irradiation process for cracking the present PSi flakes. In other words, the present PSi flakes have a very low threshold to trigger their own cracking because their unique thermal properties caused low dimensionality, and high energy processes such as pulsed laser irradiation is not needed to form Si QDs from PSi flakes. This is common in the formation of Si QD samples by using an easy ultrasonification process.²⁵ 

The HF concentration of the preparation solvent mixed in the organic solution also affects the transformation yield similar to the previous laser irradiation process.⁴⁴ The inset of Fig. 6 shows the transformation yield as a function of the HF concentration. As shown in the figure, a higher HF concentration causes more efficient transformation of PSi. This is due to its higher etching rate of surface oxides on the PSi flakes and resultant promotion of the cracking effect. In fact, crystalline silicon has lower crack-resistance energy (3 J/m²), which is a good indicator for the material toughness, than silicon dioxide (8 J/m²).⁴⁵ In addition, HF induced modification to the hydride surface also promotes the hydrosilylation reaction. The mechanical stress induced by such a chemical reaction may cause the cracking of PSi flakes. This is consistent with the previous reports in which Si QDs are formed by refluxing PSi in the mixture of toluene and alkene where a hydrosilylation reaction occurs.^46,47

To confirm the validity of the thermally induced cracking model, we estimated the activation energy of the morphology transformation from the data in Fig. 6. The solid curve in the figure represents the fit-determined curve based on the Arrhenius law, Φ = Aexp[−E_a/RT], where Φ, E_a, R, and T are the transformation yield, activation energy, molar gas constant, and absolute temperature, respectively. The fit obtained activation energy is E_a = 1.2 kJ/g. By multiplying the crack resistant energy of bulk Si (3 J/m²) by the typical surface area of PSi (100 m²/g–800 m²/g),⁴⁸ we obtain the energy to crack PSi as 0.3 kJ/g–2.4 kJ/g. Interestingly, the experimentally determined activation energy (1.2 kJ/g) falls within the calculated energy range. Although the crack resistant energy is derived from a very simplified geometrical model⁴⁵ and calculated energy value cannot directly compare with the activation energy for cracking in the PSi having complicated structure, the similar order of magnitude between them supports that the low-temperature thermal cracking occurs owing to the peculiar geometrical structure of PSi.

Finally, we discuss the potential of the present moderate thermal process for the mass production of colloidal Si QDs from PSi. When we use 5 mg–10 mg of PSi flakes per experimental batch, it has been confirmed that the yields of QDs are independent of the amount of PSi. However, in the case of the larger amount of PSi flakes (50 mg) used for the purpose of mass production per batch, the heat treatment with a larger amount of reaction solvents at 50 °C only generates ∼30 mg of Si QDs with a waste of raw flakes, i.e., 60% of transformation yield. An appropriate balance between the amount of raw material (PSi) and reaction solvent is required for further step toward the mass production of Si QDs without waste.

In Table I, we summarize the amount of Si QD yield obtained from some typical scalable production processes. As seen in this table, the amount of Si QDs per batch by using the present process is larger than the popular top-down process, i.e., laser ablation in liquid (1.5 mg),²² and the present amount is close to that of a high yield plasma synthesis method (200 mg), where oxide covered red-emitting Si QDs were prepared.¹² In the solution-based chemical synthesis processes by Zhong et al.⁴⁹ and Bose et al.,⁵⁰ the transformation yield is much lower than the process presented here. Furthermore, the properties of these Si QDs are clearly different, i.e., prepared Si QDs exhibit the PL emission not in the red region, but in the blue or green region. It should be noted that the red emission in our Si QDs is attributed to the electron–hole recombination at size dependent indirect bandgap because the Si QDs were produced from the fragmentation of PSi cores and have similar emission energies to original PSi as mentioned above. On the other hand, various emission origins for blue to green emission of colloidal Si QDs have been proposed, e.g., impurity-induced surface states^51,52 and chemical species.⁵³ 

We demonstrated the waste-free transformation of PSi flakes lifted from electrochemically etched silicon wafer into colloidal Si QDs via moderate heat treatment in a mixture of organic and HF aqueous solutions. The as-prepared Si QDs with organically passivated surfaces exhibited brighter PL emission than original PSi. The dispersibility of the Si QDs in solvents can be controlled by using organic solutions with appropriate function groups. Under an optimum condition, complete transformation from PSi into Si QDs was attained owing to the separation of individual Si cores via thermally induced fragmentation of flakes due to the unique thermal properties of PSi. The present top-down green process promising a high yield is significantly energy- and cost-effective compared to the conventional approaches. High controllability of surface termination and efficient luminescence are also excellent advantages for the present process. This self-regulated practical process can open the possibility for future scalable use of Si QDs as a printable material in optoelectronic and biological applications.

See the supplementary material for the comparison of PL spectral shape between colloidal Si QDs prepared by moderate heat treatment and UV pulsed laser irradiation processes.

The data that support the findings of this study are available from the corresponding author upon reasonable request.

The authors would like to thank Dr. Shimada (Quantum 14 KK Inc.) for the experimental support and useful discussions.