II. OPTICAL PROPERTIES OF AS-PREPARED AND OXIDIZED Si-nc

As anticipated in the Sec. I, the origin of the PL emission from Si-nc has been discussed for years: the controversy has focused on whether the photons arise from localized defect statesReference Prokes, Carlos and Glembocki^8, Reference Godefroo, Hayne, Jivanescu, Stesmans, Zacharias, Lebedev, Van Tendeloo and Moshchalkov³⁹ or the emission is ruled by the quantum confinement (QC) effect.Reference Delarue, Allan and Lannoo^34, Reference Godefroo, Hayne, Jivanescu, Stesmans, Zacharias, Lebedev, Van Tendeloo and Moshchalkov³⁹ There is now a general consensus on the possibility that both defect-mediated and e–h quantum-confined recombination mechanisms play a role,Reference Wolkin, Jorne, Fauchet, Allan and Delarue³⁵ but a universally accepted model for the origin of the PL emission in Si-nc is still lacking.Reference Goesele⁴⁵ In general, all the studies are carried out on materials fabricated by very different methods, and consequently, comparison of published results is not unambiguous. To overcome this kind of uncertainties in assessing the role of oxide states in the PL emission, here, we compare the spectroscopic properties of the same batch of crystalline Si nanoparticles, as prepared by laser pyrolysisReference D’Amato, Falconieri, Fabbri, Bello and Borsella¹²and after complete conversion to amorphous silica by alkali etching-assisted oxidation procedure.Reference Borsella, Falconieri, Herlin, Loschenov, Miserocchi, Nie, Rivolta, Ryabova, Wang, Pavesi and Turan^24, Reference Borsella, D’Amato, Fabbri, Falconieri, Trave, Bello, Mattei, Nie and Wang⁴⁶

Si nanoparticles have been synthesized by CO₂ laser-driven pyrolysis process in a continuous gas-flow reactor using silane (SiH₄) as gas precursor. The experimental setup and the procedures for laser synthesis of Si-nc with variable size are described in detail elsewhere.Reference D’Amato, Falconieri, Fabbri, Bello and Borsella¹² Briefly, silane is heated by resonant absorption of CO₂laser photons (at about 10 μm) in a small well-defined irradiation volume up to molecular dissociation, followed by growth of Si nanoparticles in a very short time (few milliseconds) by coagulation and coalescence of reactive radicals. The growth is abruptly terminated as soon as the particles leave the irradiation region, leading to an average particle size in the nanometer scale (5–15 nm) and a narrow size distribution. The final particle size can be controlled acting on the process parameters (i.e., total pressure, gas flow rates, etc.) to change the residence time of the particles in the reaction zone.

The as-prepared Si nanoparticles emit in the range 550–950 nm under excitation at wave lengths below 500 nm. The luminescence excitation cross section of Si-nc obtained by laser pyrolysis were previously measured with Ar laser excitation at 488 nm, using the luminescence rise time method,Reference Borsella, D’Amato, Fabbri, Falconieri, Trave, Bello, Mattei, Nie and Wang⁴⁶ giving values in the range of 1.5 ÷ 8 × 10⁻¹⁶ cm². Typical PL emission quantum yield was estimated to be of the order of 1% by use of a procedure based on a dye standard (rhodamine 6G).

The nanoparticles synthesized for the present study are composed by Si-nc with diameter d = 5.6 nm, narrow size distribution, and capped by a thin oxide layer (≤1 nm), as determined by high-resolution transmission electron microscope (HR-TEM) analysis. A stable suspension of Si-nc in methanol is obtained by high power ultrasonic treatment. The liquid is opaque and brown yellowish in color. A second sample is prepared by submitting the Si nanoparticles to a strong alkali etching procedure, stopped by addition of H₂O. The complete oxidation of the sample is evidenced by the compositional analysis giving an atomic ratio O(k)/Si(k) = 1.9 on different portions of the sample and by the HR-TEM analysis confirming the formation of amorphous nanoparticles of few nanometer size. The suspension of amorphous silica nanoparticles (SiO₂-np) in water is colorless.

Room temperature continuous wave (CW) photoluminescence emission and excitation (PLE) spectra measurements were performed using a HORIBA-Jobin-Yvon Fluorolog 3-22 spectrofluorometer (HORIBA Instruments Incorporated, Edison, NJ). Excitation spectra were collected from diluted samples placed in a thin (0.1 cm) cuvette to fulfill the optically thin sample condition.

Both the Si-nc and the SiO₂-np samples show very similar broad PL spectra in the range 550–950 nm under excitation at 355 nm (Fig. 1, inset), the only difference being a small blueshift of the oxidized sample with respect to the Si-nc sample. On the other hand, the difference in material composition and electronic structure of the two samples is quite evident in the PLE spectra reported in Fig. 1. In the limit of optically thin sample, the PLE signals are proportional to the absorption and can be interpreted on the basis of the expected spectral dependence of the absorption coefficient α of the two different materials.

FIG. 1.

CW PL excitation and emission spectra (inset) of Si-nc (filled circles) and SiO₂-np samples (empty circles). Dotted lines show the decomposition of PLE spectrum of SiO₂-np sample in two Gaussians as described in the text.

In the case of Si-nc, at low excitation energies ≤2.3 eV, the PLE signal (α_PLE) follows the behavior of indirect transitions.Reference Kovalev, Heckler, Polisski and Koch^44, Reference Ben-Chorin, Averboukh, Kovalev, Polisski and Koch⁴⁷

(1)$$<![CDATA[$$({\rm{\alpha }}_{{\rm{PLE}}} \hbar {\rm{\omega }})^{0.5} \, = \,A_1 (\,\hbar {\rm{\omega }} - E_{g1} )\quad.$$

The E_{g 1} gap extrapolated by fitting with the formula [Eq. (1)], the experimental curve of Fig. 2 is ∼1.84 eV, in relatively good accordance with the value of 1.9 eV reported by KovalevReference Kovalev, Heckler, Polisski and Koch⁴⁴ and Ben-Chorin et al.Reference Ben-Chorin, Averboukh, Kovalev, Polisski and Koch⁴⁷ These authors associate the E_{g 1} gap with the Γ_25′,V –Δ_1,C transition in Si. Above 2.3 eV, a deviation from Eq. (1)is observed. The extra absorption in this energy region follows a similar dependence:

(2)$$<![CDATA[$$({\rm{\alpha }}_{{\rm{PLE}}} \hbar {\rm{\omega }})^{0.5} \, = \,A_2 (\hbar {\rm{\omega }} - E_{g2} )\quad,$$

FIG. 2.

CW PL excitation spectrum of Si-nc (filled circles) and analysis showing the two indirect band gaps and the direct band gap (inset).

but with a different energy gap E_{g 2}, which results to be 2.03 eV (whereas a value of 2.5 eV is found in Refs. Reference Kovalev, Heckler, Polisski and Koch44 and Reference Ben-Chorin, Averboukh, Kovalev, Polisski and Koch47). As expected from the QC theory, this value is blueshifted with respect to 1.6 eV value found for crystalline bulk Si and attributed to the Γ_25′,V –L_1,Ctransition.Reference Kovalev, Heckler, Polisski and Koch^44, Reference Ben-Chorin, Averboukh, Kovalev, Polisski and Koch⁴⁷

At ħω ≥ 3.2 eV, there is a further change in the slope, due to the occurrence of direct transitions.Reference Kovalev, Heckler, Polisski and Koch^44, Reference Ben-Chorin, Averboukh, Kovalev, Polisski and Koch⁴⁷The analysis of PLE data shows that the absorption of Si-nc has a similar dependence to that of bulk Si. In conclusion, the main effect of QC seems to be the blueshifting of transitions associated with different critical points of the band structure.

On the other hand, the PLE spectrum of the oxidized amorphous sample (SiO₂-np) is totally different from that of the nanocrystalline Si-nc sample (see Fig. 1). A tentative interpretation is based on the comparison with absorption features observed in similar systems. It is well known that silica is an insulator with a band gap of about 11 eV; however, PL emission from defects and impurities in silica can be induced by ultraviolet (UV)–visible light absorption.Reference Skuja^48, Reference Munekuni, Yamanaka, Shimogaichi, Tohmon, Ohki, Nagasawa and Hama⁴⁹ In particular, an emission band peaked at 652.6 nm (1.9 eV) is observed after absorption at about 2.0, 4.8, and 7.6 eV in defective silica.Reference Skuja^48, Reference Munekuni, Yamanaka, Shimogaichi, Tohmon, Ohki, Nagasawa and Hama⁴⁹ The absorption at 4.8 eV is attributed to a transition from the σ bonding orbital to the 2p nonbonding orbital of nonbridging oxygens in defects commonly labeled as NBOHC. The NBOHC is an under-coordinated oxygen with a hole in its nonbonding 2p orbital.Reference Skuja^48, Reference Munekuni, Yamanaka, Shimogaichi, Tohmon, Ohki, Nagasawa and Hama⁴⁹ The resulting red PL emission (R-band) is due to the charge transfer transition between the nonbonding 2p orbital of the nonbridging oxygen and the lone pair 2p orbital of one of the nearest bridging oxygens.Reference Skuja⁴⁸ More recently, a broad red NIR PL emission band from mesoporous silica was detected in the range 600–1040 nm under excitation at about 4.7, 3.5, and 2.3 eV.Reference Glinka, Lin, Hwang and Chen^37, Reference Glinka, Zyubin, Mobel, Lin, Hwang and Chen^38, Reference Zyubin, Glinka, Mebel, Lin, Hwang and Chen⁵⁰ Mesoporous silica is characterized by pores with sizes in the range 2–10 nm and the thickness of walls among pores is ∼1 nm, which is of the same range of the oxide layer around Si-nc.Reference D’Amato, Falconieri, Fabbri, Bello and Borsella¹²The PLE spectrum of the oxidized sample was compatible with the presence of two broad Gaussian absorption bands: the first centered at about 4.8 eV and the second at much higher energy, located beyond our accessible energy range. Therefore, the PLE data reasonably confirm the defective silica composition of this sample.

The strong similarity of the luminescence emissions of the two materials is also observed in the luminescence time decay data detected in the whole 550–950 nm range under excitation with a frequency-tripled pulsed Nd-YAG laser emitting at 355 nm. To facilitate the comparison, data taken at some selected emission wave lengths on both samples are reported in Fig. 3. It is fairly evident that the decay curves are not single exponentials and show increasingly fast decay at shorter detection wave lengths. Having assumed that the PL emission of the oxidized sample comes from defects in amorphous silica, its time decay can be described by a sum of single exponential terms, each one representing the emission from a specific radiative defect. In particular, we fitted the experimental data using a function of the form:

(3)$$<![CDATA[$$I({\rm{\lambda }},t)\, = \,\mathop\sum\limits_i {A_i ({\rm{\lambda }})\exp ( - t/{\rm{\tau }}_i )\, + \,y_0 ({\rm{\lambda }})\quad,} $$

FIG. 3.

Luminescence time decay curves of Si-nc (filled symbols) and of fully oxidized (empty symbols) samples at some selected emission wave lengths (squares: λ = 670 nm, circles: λ = 770 nm, triangles: λ = 925 nm). Continuous lines represent the theoretical fits as described in the text.

where the $A_i ({\rm{\lambda }})$ coefficients represent the spectral distribution of the emission of the ith defect center, with decay time τ_i.

The set of time decay data of the oxidized sample was fitted with the function I(λ,t) [Eq. (3)], adding progressively more exponential terms until the χ² factor of the fit was satisfactory. It was found that four exponential terms are necessary and sufficient to obtain a good fit of the set of decay curves of the oxidized sample, as shown in Fig. 3. The values of the time constants obtained from the fit are reported in Table I. The strong similarity of the spectral and time dependence of the PL emission from both the Si-nc and oxidized samples (see Figs. 1 and 3) led us to hypothesize that also in the former sample, the emission can be due to radiative defects. Therefore, the same fitting procedure based on Eq. (3) was used to find the necessary number of exponential terms to fit the decay times. As shown in Fig. 3, also for Si-nc, the sum of four exponential terms is adequate to reproduce the experimental time decays. The time constants are reported in Table I.

TABLE I.

Time constants obtained from the multiexponential fitting of the PL time decay data of Si-nc and SiO₂-np samples.

It should be noted that luminescence time decays from Si-nc are usually analyzed in terms of the stretched exponential function,Reference Pavesi and Ceschini⁵¹ even in systems where this model is not fully justified. On the other hand, multiexponential analysis of PL decay curves from Si-nc produced by Si implantation in silica was also adopted in Ref. Reference Jayatilleka, Diamare, Wojdak, Kenyon, Mokry, Simpson, Knights, Crowe and Halsall52, but the exponential terms were attributed to emission from different nanocrystal classes. In that paper, three time constants were determined, whose values, 220, 75, and 50 μs, compare relatively well with ours, considering the very different material examined.

The time constants reported in Table I are remarkably similar for both samples confirming our hypothesis that the same emitting centers are present in both systems. To get further insight into the physical meaning of our proposed analysis, it must be recalled that the $A_i ({\rm{\lambda }})$ coefficients in Eq. (3) represent the spectral distribution of the emission of the ith center. It should then be possible to reproduce the CW PL emission spectra shown in the inset of Fig. 1by use of the following formula:

(4)$$<![CDATA[$$I(E)\, \propto\, \mathop\sum\limits_i {{\rm{\tau }}_i A_i (E)\quad.} $$

To this purpose, after conversion to the energy scale, the A_i(E) coefficients were fitted with Gaussian functions as appropriate for inhomogeneously broadened emitters. The center energies of the four Gaussians are reported in Table IIfor both samples.

TABLE II.

Center energies (in eV) of the Gaussian functions that best fit the energy dependence of the Ai coefficients.

After having been multiplied by the corresponding τ_i and a normalization factor, both the coefficients A_i(E) and the Gaussian fits have been reported in Fig. 4together with the CW PL emission spectra of both the samples.

FIG. 4.

Amplitudes of the single exponential components as derived from analysis of PL decay curves (empty symbols) and their fits with Gaussian functions (lines), for Si-nc (a) and SiO₂-np (b) samples. Insets: close-up of the components peaked at the highest energy for both materials. Also shown are the PL emission spectra of the two samples (dotted lines) and their reconstruction with the sum of the Gaussian components (filled symbols).

The satisfactory reproduction (see Fig. 4) of the PL emission spectra of both samples with four Gaussian functions having slightly shifted centers (Table II) and the similarity of the corresponding decay times (Table I) pushes toward the identification of the same type of emitting centers as responsible for the PL emission in both the Si-nc and the fully oxidized samples. To this purpose, we can rely on the huge amount of literature on the luminescence from defective silica and in particular on the role of the NBOHC centers. PL emission from these defects has been formerly reported to be at the origin of the red luminescence band in the range 1.8–1.9 eV (≈690–650 nm) in bulk silica,Reference Skuja⁴⁸ while the presence of surface NBOHC centers is responsible for the appearance of two subbands at 1.93 and 1.98 eV. At least, three different types of red-emitting NBOHC were identified by Munekuni et al.Reference Munekuni, Yamanaka, Shimogaichi, Tohmon, Ohki, Nagasawa and Hama⁴⁹ in silica glass after γ-ray irradiation and were related predominantly to the change in hydride or hydroxyl content. Accordingly, a sample-to-sample variation in the PL band was observed depending on the relative numbers of the different emitting defects. Successively, in nanosized and mesoporous silica, Glinka et al.Reference Glinka, Zyubin, Mobel, Lin, Hwang and Chen^38, Reference Zyubin, Glinka, Mebel, Lin, Hwang and Chen⁵⁰have observed PL bands peaked at 1.90 eV (652.6 nm), 1.78 eV (696. nm), 1.61 eV (770.2 nm), and 1.40 eV (885.7 nm), and smaller contributions at 1.27 eV (976.4 nm) and 1.14 eV (1087.7 nm), with decay times compatible with those observed in our samples. Ab initio modeling has confirmed the strong sensitivity of the energy levels of NBOHC to the atomic surrounding.Reference Zyubin, Glinka, Mebel, Lin, Hwang and Chen⁵⁰

In agreement with data shown in Fig. 1 and with literature data on NBOHC absorption bands,Reference Skuja^48, Reference Munekuni, Yamanaka, Shimogaichi, Tohmon, Ohki, Nagasawa and Hama⁴⁹ excitation of the NBOHC defects in the fully oxidized sample is assumed to proceed via the broad absorption bands at 4.8 and 7.6 eV. On the other hand, in the as-prepared sample, the PLE data clearly show the features of the electronic absorption of Si-nc, as previously discussed and shown in Fig. 2. Therefore, it is quite reasonable to postulate trapping of the photoexcited Si-nc carriers in the luminescent defects. This process can take place via the lowest and weaker NBOHC absorption band at about 2 eV because this energy matches well the energy of relaxed carriers in the Si-nc (see the band gap analysis reported in Fig. 2 and the scheme in Fig. 5). Radiation is then emitted as a consequence of charge transfer from the nonbonding oxygen orbital to the 2p orbital of one of the three other oxygen ligands. As stressed in Ref. Reference Zyubin, Glinka, Mebel, Lin, Hwang and Chen50, both strain and chemistry must be taken into account when predicting the range of the emission energies.

FIG. 5.

Scheme of the non-radiative deexcitation processes from photoexcited Si-nc either to interface defect states or to nearby rare-earth ions.

III. OPTICAL PROPERTIES OF Yb-DOPED Si-nc

According to the considerations made above, it is clear that light excitation of the Si-nc prepared for this study is affected by the QC effect, but the broad PL emission band detected in the 550–950 nm range is due to contributions from several radiative defects and cannot be straightforwardly tuned through the control of the nanocrystalline size. Moreover, slight variations in the preparation and processing routes of the Si-nc can lead to uncontrollable changes in the emission intensity and spectral features.Reference Wang, Ni, Pietzsch, Hennies, Bao and Chao¹⁹ On these premises, we tried a different strategy to attain specific colored emission in an engineered fluorophore. As anticipated in the Sec. I, we experimented the application of pyrolytic Si-nc as antennas to absorb excitation light and successively transfer the energy to a rare-earth emitter placed nearby the nanocrystal. Yb³⁺ ions were chosen as prototype emitters because they represent a suitable case study, as detailed in the following. First of all, their extremely simple energy level structure does not permit direct excitation at energies above the luminescent level as shown in Fig. 5; in this way, occurrence of energy transfer from the sensitizer can be readily assessed. Second, their emission band at 980 nm is such to keep the possibility of detection with inexpensive silicon detectors, and third, Yb³⁺ions can be linked to Si-nc oxide surface exploiting their ability to form complexes with amino groups.

In our procedure, Si-nc prepared by laser pyrolysis (with crystallite size equal to 9 nm as determined from x-ray diffraction analysis) were first functionalized with amino groups by reaction with aminopropyltriethoxysilane (APTES) and subsequently dispersed in a solution of EtOH and YbCl₃ to induce the reaction of Yb-ions with the surface-bound NH₂ groups. The complex [Yb(III)(APTES)₃Cl₃] was also synthesized to check possible unwanted effects of APTES residuals on the optical properties of the functionalized Si-nc. Yb-doped Si-nc were annealed at 300° C for 30 min in air to remove residual alcohol groups, which could quench the Yb-ion PL emission.

The FTIR spectra of the as-prepared Si nanoparticles and after functionalization either with APTES only or with both APTES and Yb-ions are shown in Fig. 6. The FTIR spectrum of the as-prepared sample shows signals due to silicon bonds with hydrogen and oxygen, while the ammine groups are clearly visible in the spectra taken on the functionalized samples. Moreover, the ammine band broadens after Yb doping, which strengthens the assumption of effective complex formation.

FIG. 6.

FTIR spectra of as-prepared Si-nc, NH₂-functionalized Si-nc, and Yb-doped Si-nc.

The as-prepared Si-nc do not show any detectable luminescence emission, probably as a result of poor surface oxidation. On the other hand, PL emission from Yb-ions is clearly detected under laser excitation at 532 nm of the Yb-doped Si-nc samples. Results are reported in Fig. 7. As expected, the emission intensity approximately scales up with the Yb content in the samples prepared with two different YbCl₃ concentrations. It was checked that the PL signal is absent after excitation at 532 nm of the reference complex [Yb(III)(APTES)₃Cl₃], thus ruling out a possible sensitization effect of APTES on the Yb-ions. The sensitization effect of Si-nc is then unambiguous since the electronic structure of Yb³⁺ ions does not permit direct excitation in the visible. To further prove the broadband sensitization of the Si-nc, the excitation spectrum of the Yb-ions emission at 980 nm was measured and is reported in the inset of Fig. 7. The similarity with the typical PLE spectra of luminescent Si-nc confirms the nanocrystals sensitizing role.

FIG. 7.

CW PL spectra of Yb-doped Si-nc at two different Si-nc:Yb molar ratios, excited at 532 nm. Inset: excitation spectra of Yb emission at 975 nm.

Our strategy of doping Si-nc with rare-earth ions can be extended to Si-nc obtained with other preparation methods since it offers the possibility to get well controllable, narrow band, and stable PL emission.

IV. TOXICITY TESTS

Due to the considerable interest for the use of silicon nanoparticles in a medical context,Reference Wang, Ni, Pietzsch, Hennies, Bao and Chao^19, Reference Choi, Wang and Reipa^40, Reference Tasciotti, Liu, Bhavane, Plant, Leonard, Price, Cheng, Decuzzi, Tour, Robertson and Ferrari⁵³ here, we report some specific in vitro studies that represent a relatively rapid and cost-effective way of screening in the assessment of nanoparticles toxicity. The biocompatibility of Si nanoparticles was estimated by assessing the integrity of the plasma membrane, by means of LDH assay, following the incubation of the cells with nanoparticles for 6 or 24 h. Among the biocompatibility assays, the LDH assay is considered one of the least sensitive since it evaluates a final event in a cell life, the rupture of the plasma membrane. For this reason, we performed also the MTT assay, which was the most sensitive cytotoxicity assay together with the neutral red one.Reference Fotakis and Timbrell⁵⁴ The LDH release assay doses the activity of enzyme LDH in the extracellular medium. LDH is a cytoplasmic enzyme that should not be present outside its normal microenvironment. Its activity detected outer the cytosol, thus in the culture medium indicates an ultrastructural plasma membrane damage. The MTT assay was used to test the cell viability by probing the mitochondrial activity on cells incubated for 24 h. MTT is a water-soluble tetrazolium salt, which is converted to an insoluble purple formazan by cleavage of the tetrazolium ring by the mitochondrial enzyme succinate dehydrogenase. The formazan product is membrane impermeable, thus it accumulates in healthy cells. In the case of diagnostic imaging, Si nanoparticles are usually injected intravenously and their impact is on the endothelial wall. However, in evaluating exposure risks, it must be considered also the possible inhalation of nanoparticles that may occur in professional working conditions. In this last case, the most likely route of entry is across the lung alveolar epithelium whose surface area averages 70 m².

For biocompatibility studies, we have chosen alveolar epithelial cells (A549) and murine fibroblasts (L929). These cells, A549 and L929, were grown on Petri dishes in Dulbecco's modified Eagle's medium or Eagle's Minimum Essential medium, respectively, supplied with 10% fetal bovine serum (FBS), 1% of L-glutamine, and 1% of penicillin/streptomycin and incubated in a controlled environment at 37 °C with 5% CO₂. During all the experiments, cells were incubated with medium supplemented with 1% FBS to prevent serum interference in the toxicity assay. LDH release test (Roche Applied Science Monza, Italy) and MTT assay (Sigma Aldrich, Milano, Italy) were performed according to the manufacturer instructions. Si-nc were dissolved at different concentrations (from 1 ng/mL up to 10 μg/mL) in the cell culture medium and cells were incubated for 6 or 24 h. In some cases, after 24 h, nanoparticles were removed and fresh medium was added for 24 h to cell cultures to assess the ability of the cells to recover after nanoparticles exposure. The maximal value of LDH release was obtained on lysed cells.

Toxicity tests were performed on different batches of Si-nc prepared by laser pyrolysis.Reference D’Amato, Falconieri, Fabbri, Bello and Borsella¹² As-prepared Si-nc with average size of 8.9 nm (labeled Si193H) were dispersed in dimethyl sulfoxide in a stock concentration of 1 g/L. Since DMSO is a well-known toxic solvent, before performing nanoparticles biocompatibility tests, we determined that 0.001% is the maximum concentration of DMSO that avoids its toxicity per se, as shown in Fig. 8. Cells exposed to 0.001% DMSO in the medium revealed no signs of distress [e.g., detachment from the plate; see Fig. 8(b)], and the percentage of dead cells was not different from the control condition [see Fig. 8(a)] (9.7% ± 3% versus 5.0% ± 2%, respectively, evaluated with Trypan blue). This dilution of the solvent corresponds to nanoparticles concentration of 1 μg/mL. The percentage of LDH release was very high for the maximal concentration of nanoparticles tested (83% ± 3% and 98% ± 2% for 6 and 24 h of incubation, respectively). It has to be considered though that at this maximal concentration, the effect of the solvent is still present. In fact, at the 10 times lower concentration (100 ng/mL), the LDH release decreased significantly, being 1% ± 4% at 6 h and 38% ± 2% at 24 h of incubation [Fig. 9(a)]. Moreover, after 24 h of incubation and 24 h of recovery, the LDH release decreased to 19% ± 6%.

FIG. 8.

Three typical images acquired in bright field (magnification 4x) showing the effect of progressive concentration of DMSO on A549 cell culture. (a) Control culture and (b) and (c) DMSO 0.001% and 0.01%, respectively, in culture medium.

FIG. 9.

LDH release by A549 cells after 6 h (open squares) and 24 h (filled squares) of incubation for 8.9-nm Si-nc dispersed in DMSO (a) and 8-nm Si-nc dispersed in ethanol (b). The filled stars represent the percentage of LDH release after 24 h of nanoparticles incubation followed by 24 h of recovery in regular medium.

Other batches of Si-nc (with average size of about 8 nm) were dispersed in ethanol with a stock concentration equal to 8 g/L, almost 10 times higher than the previous one. For these nanoparticles (labeled as Si276N1 and Si276N2), at the concentration tested, we did not find any effect of the solvent (data not reported here), and also, we did not find a real dose dependence of LDH release, being about 5% after 6 h of incubation for all the concentration tested and about 30% after 24 h as shown in Fig. 9(b). Also in this case, the recovery after 24 h showed a trend, even if not pronounced, indicating a decrease in the level of the cell population membrane damage (11% ± 14%).

Moreover, we tested surface-oxidized silicon nanoparticles (S276G-F with average size of 8 nm and S252D with average size of 4.5 nm). Procedure for the chemical surface oxidation is described in Ref. Reference D’Amato, Falconieri, Fabbri, Bello and Borsella12. In this case, the incubation of A549 with nanoparticles revealed a dose dependence but not a time dependence in the LDH release, as shown in Fig. 10(a). The maximal release observed corresponded to the 15% of the highest value, both after 6 or 24 h of incubation. Since this was the case when the alveolar epithelial cells were less stressed, we tested the same Si-nc also on murine fibroblasts. As can be appreciated in Fig. 10(b), the data revealed indeed a dose-dependent plasma membrane distress with the maximal value of LDH release equal to about 30% at the maximal concentration tested. On the other hand, we did not observe any clear size-dependent effect. Finally, we tested Yb-doped Si-nc on both the alveolar epithelial and fibroblast cells after 24 h of exposure. We selected this interval of incubation since we did not detect any time dependency on oxidized Si-nc. The results showed again a dose response effect, with a maximum release of about 20% in both the cell lines (Fig. 10).

FIG. 10.

LDH release by A549 cells [panel (a)] and L929 cells [panel (b)] after 6 h (open squares) and 24 h (filled squares) of incubation with oxidized Si nanoparticles and after 24 h incubation with Yb-doped Si-nc (filled triangles) dispersed in ethanol.

The results of the MTT assays obtained with both A549 and L929 cells incubated for 24 h with as-prepared, oxidized, and Yb-doped Si-nc (Fig. 11) display the same degree of sensitivity toward the three types of Si-nc, showing a dose-dependent MTT reduction with a maximum value equal to about 30%. These data do not show significant differences with respect to the LDH assays, confirming the relatively low dose-dependent toxicity of all the tested silicon-based nanoparticles.

FIG. 11.

MTT assay of A549 [panel (a)] and of L929 [panel (b)] cells viability after 24 h incubation with as-prepared Si-nc (filled squares), oxidized Si-nc (filled circles), and Yb-doped Si-nc (filled triangles).

The literature reports a wide variety of cell lines tested and assays to evaluate the geno- or cytotoxicity. The most common tests used to study the in vitro toxicity evaluate cell morphology, apoptosis, and cells viability but also cytokine production as an indirect marker of inflammation and oxidative stress. However, our two tests provide clear assessment of cells response to Si-nc and allow a comparison with other cytotoxicity data reported in the literature. Our data suggest that, in general, Si-based nanoparticles per se do not reveal a high level of cells distress. Moreover, when we exposed the cell culture to nanoparticles for 24 h and then let the cells recover substituting the medium containing nanoparticles with fresh medium for other 24 h, we found a lower level of LDH release. This may indicate that the greater part of the cell population is able to recover from the nanoparticle interaction.

To investigate the behavior of different cell lines, we tested the nanoparticles that were less reactive on A549 cells also on the murine fibroblasts L929. These cells are widely distributed in the skin and constitute a common component of the connective tissue whose role is to secrete precursors of the extracellular matrix such as collagen and proteoglycan. They are metabolically active and in fact are used as tool in many standard material biocompatibility tests.Reference Zange, Li and Kissel^55, Reference Nordin, Wieslander, Martinson and Kjellstrand⁵⁶ Our results suggest that fibroblasts are more sensitive than epithelial cells, confirming that not all the cell lines behave at the same and that for a more thorough analysis, one has to consider either the specific biological target or a pool of cell lines representative of the whole organism. A similar conclusion was drawn by Fujioka et al.Reference Fujioka, Hiruoka, Sato, Manabe, Myasaka, Hanada, Hoshimo, Tilley, Manome, Hirakuri and Yamamoto⁵⁷ in a recent paper where they compared the sensitivity of HeLa cells and hepatocytes to CdSe QD. In this same paper, studying the response of HeLa cells to luminescent oxidized Si-nc in terms of mitochondrial activity assays and LDH release assays, they showed that the Si-nc are safer than CdSe quantum dots, especially under UV exposure conditions and below the concentration of 112 μg/mL. Our work reports data obtained incubating the cells well below that threshold, and still, we found a moderate membrane distress that, considering again the work of Fujioka et al.,Reference Fujioka, Hiruoka, Sato, Manabe, Myasaka, Hanada, Hoshimo, Tilley, Manome, Hirakuri and Yamamoto⁵⁷ might be the consequence of oxygen radicals production affecting the plasma membrane of the cells. Considering also our results obtained investigating the mitochondrial distress, we find only a minor deviation from the data presented in the literature,Reference Wang, Ni, Pietzsch, Hennies, Bao and Chao^19, Reference Erogbogbo, Yong, Roy, Xu, Prasad and Swihart^22, Reference Erogbogbo, Yong, Roy, Hu, Law, Zhao, Ding, Wu, Kumar, Swihart and Prasad²⁷ thus confirming that for a concentration in the range of μg/mL, Si-nc can be considered biocompatible. Finally, in contrast with Bhattacharjee et al.,Reference Bhattacharjee, de Haan, Evers, Jiang, Marcelis, Zuilhof, Rietjens and Alink⁵⁸ it is noteworthy to observe that we do not detect any difference in MTT production despite the different charge of the Si-nc. In fact, the measured Z-potential was negative for the as-prepared nanoparticles and for the oxidized nanoparticles (−32.8 ± 1.1 mV and −17.6 ± 0.9 mV, respectively), while it was positive for the Yb-doped nanoparticles (+35.9 ± 1.1 mV).