Nuclear magnetic resonance

While the presence of the products of oxidation is negligible in the solvent itself^¶, its importance clearly rises with the UV-irradiation procedure (i.e. in the reference irra-diated xylene and the Si-coll), suggesting that the UV irradiation promotes oxidation and, moreover, may serve as a source of energy for this reaction. Furthermore, the shape of the (gray in Fig. 3.5d) OH band, being broad and asymmetric, is a clear fin-gerprint of carboxylic acids, implying that the oxidation reaction proposed in Eq. 3.3 proceeds as far as its third stage.

Another sign pointing toward UV-irradiation-triggered oxidation is the presence of the (red in Fig. 3.5d) band around ∼1265 cm⁻¹ in the irradiated samples (both reference Xand colloidal 1sedX), whose identification, however, is not straightforward.

Most probably, it is connected either with an aromatic aldehyde and/or dicarboxylic (phthalic) acids. When comparing the 1sedX and reference X samples, it is obvious that both contain the oxidation products, but the oxidation seems to be more intense, gaining a higher concentration of aldehydes and carboxylic acids, in 1sedX, in which even the carboxylic-acid-characteristic (orange in Fig. 3.5d) overtones around 2550 and 2660 cm⁻¹ are clearly distinguishable.

Besides the oxidation-products-related compounds, however, the spectrum of 1sedX contains one more very important band at∼1280 cm⁻¹, unfortunately located very near to the band at∼1265 cm⁻¹ (it can be more easily discerned in a zoomed-in comparison of the acquired FTIR-ATR spectra in Fig. 3.5c). This band, according to Tab. 3.1a, can be identified with Si-CH_n bonds and should have a counterpart between 765 and 870 cm⁻¹. Although this interval significantly overlaps with the vibrations of xylenes, a new small band appears in1sedXat 790 cm⁻¹ (Fig. 3.5cin green). Since it is different from the bands in other samples and, moreover, it gets more intense while the solute concentrates in the solvent during the measurement, we believe that the combination of 790-cm⁻¹ and 1280-cm⁻¹ bands can be ascribed to the Si-CH_n vibration.

The appearance of Si-CH_n bond in the 1sedX sample means that the (by)products of oxidation are at the final stage bonded directly to silicon, presumably forming an organic capping on the surface of the SiNc. In this context, the comparison of the original oxide surface passivation (in dark blue in Fig. 3.5d) of both the pure SiNc powder and the powder dispersed in xylene prior to the UV-irradiation treatment with 1sedX is well worth mentioning. While the silicon oxide Si-O-Si vibrations are the dominant feature in the spectra of both the non-UV-irradiated samples, they are strongly suppressed in the spectrum of the resulting colloid1sedX^k, suggesting that the organic capping actually replaces the original oxide passivation.

3.2. Surface chemistry probing the splitting of energetic spin sublevels of nuclei with a non-zero spin (most commonly ¹H and ¹³C) in an applied external magnetic field with a radio-frequency electromagnetic field: whenever the frequency of the probing field resonates with the magnetic-field-induced splitting, absorption of the probing field is observed. Impor-tantly, the characteristic frequency depends not only on the characteristic frequency of the studied nucleus^∗, but also on the shielding effects of the surrounding electrons (the so-called chemical shift) and, eventually, on the presence of other non-equivalent nu-clei, which are up to three chemical bonds away from the studied nucleus (the so-called spin-spin or J coupling). As a result, NMR can detect and identify various chemical functional groups in the studied material.

Since both the frequency shift and the fundamental resonant frequency are directly proportional to the strength of the magnetic field, the shift is converted into a field-independent dimensionless values: the difference between the frequency of a reference and the studied sample is divided with the frequency of the reference. This quantity is typically very small, and is expressed as parts per million (ppm).

Instead of continuous-wave measurements, which would be (and were) substantially time-consuming due to low signal-to-noise ratio, Fourier-transform approach combined with pulsed excitation are being applied. In addition to an improved signal-to-noise ratio, the pulsed excitation allows for the study of the relaxation of the spin-related magnetization in the coherent regime, which can yield many different types of informa-tion. Apart from simple Fourier transformed magnetization decay, yielding traditional one-dimensional NMR spectrum, it is possible to apply a series of pulses and study the transfer of excited magnetization to neighboring atoms. Thus gained Fourier trans-formed map depicts NMR intensity as a function of both the chemical shift and e.g.

spin-spin coupling (or two different chemical shifts) and makes it possible to combine assigning of various functional groups with the identification of neighboring attached nuclei or functional groups.

The most common 2D NMR map is a heteronuclear^† single quantum coherence (HSQC) spectrum, which, in theory, should show one peak for every hydrogen bonded to the heteronucleus in the 2D map.

Besides using other chemical shift as a second dimension, it is also possible to acquire different NMR spectra for species with different diffusion coefficients (size) by applying a series of pulses which dephases the magnetization from more quickly diffusing spins but does not affect the slower ones.

NMR measurements of xylene-based samples: First of all, the composition of the pure sol-vent was investigated. It turned out that the solsol-vent contains, apart from xylene isomers, a sizable amount of ethylbenzene and isopropylbenzene^‡ (see Tab. 3.2), and probably also other compounds. The composition of the solvent corresponds most likely to the so-called technical xylene.

∗or more specifically on its gyromagnetic ratio

†meaning pertaining to two different nuclei, e.g. ¹H–¹³C

‡i.e., the solvent referred to in this thesis so far as “xylene” is in fact a mixture of xylene isomers, ethylbenzene and isopropylbenzene, for the sake of simplicity the same name will also be used from now on

compound formula solvent Xall ratio ν_evap

m-xylene 37 % 29 0.78 0.70

ethylbenzene 28 % 12 0.44 0.84

isopropyl-benzene (cumene)

14 % 6 0.43 0.50

o-xylene 9 % 5 0.55 0.72

p-xylene 12 % 12 1 0.70

Table 3.2: Chemical composition of the solvent and Si-coll. The ratio of chemical compounds in the Xall sample is calculated assuming that the contents of p-xylene (which drops off the least) does not change (absolute change cannot be measured unless a concentration standard is added to the sample).

The last column contains the values of evaporation rates with respect to a standard (n-butyl-acetate).

The same compounds were also found in the colloidal sample Xall, however, they occurred in different ratios (see again Tab. 3.2). Interestingly, the most pronounced decrease in concentration (after the stirring/UV-irradiation procedure) was observed in the case of ethylbenzene and isopropylbenzene, while the concentration of xylene isomers dropped off less. This difference cannot be simply explained by differences in evaporation rates (isopropylbenzene should evaporate the least, see last column in Tab. 3.2), therefore it is highly probable that the compounds are consumed in a chemical reaction.

As a result, we believe that we can specify the oxidation reaction from Eq. 3.3.

We propose that one of the key chemical reactions taking place during the preparation procedure of the Si-coll is the oxidation of ethylbenzene^[A1] into acetophenone and benzoic acid^§:

.

(3.4) However, NMR spectra of such a complex mixture of chemical compounds as our solvent are inevitably very complicated since they contain a large number of lines. To

§interestingly, the presence of benzoic acid is consistent with both NMR and FTIR measurements

3.2. Surface chemistry

(a) (b)

Figure 3.6: NMR measurements of 1sedXD in chloroform: (a) (after^[A1]) shows three types of ¹H NMR spectra and (b) represents a¹³C NMR spectrum. The diffusion NMR in red in (a) contains the diffusion-filtered ¹H NMR spectrum for the diffusion coefficient around 150 µm²/s, which is in fact curve in the tenth row in the whole diffusion NMR measurement shown in Fig. 3.7. Red arrows in (a) indicate persisting aromatic and aliphatic signal, green arrow in (b) highlights the position of the Si-C peak.

overcome this hindrance, we let xylene dry out from the colloid and redispersed the solute in pure deuterated chloroform CDCl₃^¶ using sample 1sedXD. The procedure is described in Sec. 2.2.3.

NMR measurements of1sedXDin chloroform: The redispersion of the solute from 1sedXD in deuterated chloroform helped us obtain clearer NMR spectra. Even if the NMR spectra still contain a complex family of resonances, which makes it impossible to assign individual lines to corresponding functional groups, we were still able to draw important conclusions based on the acquired NMR spectra.

Although the traditional ¹H NMR spectrum of 1sedXD, which is shown in black in Fig. 3.6a^[A1], consists of a large number of lines of both aliphatic and aromatic groups, suggesting that more types of similar molecules are present in the colloid, a single small line at ∼0.1 ppm (emphasized with a green band) is of particular interest. The chemical shift of this line is characteristic of a hydrogen atom chemically bond to (i.e.

2–3 chemical bonds away from) a silicon atom, suggesting the feasibility of detecting SiNcs, even it is present only in low concentration, in our colloid.

As this 0.1-ppm line does not by far represent a dominant feature of the ¹H NMR spectrum, further measurements were essential to corroborate this conclusion. Obvi-ously, NMR spectra detecting the presence of other atoms than a hydrogen one are perfect candidates for the task. However, these measurements have a very low

signal-¶the¹H NMR spectrum of deuterated chloroform CDCl₃ consists of a single sharp line which can serve as a reference

to-noise ratio since they usually rely on the natural occurrence of unstable isotopes in the sample (¹³C, ²⁹Si)^k. Consequently, only a small percentage of atoms can actually be probed and very long acquisition times are necessary.

Nevertheless, a¹³C NMR spectrum, shown inFig. 3.6b, which once again proved to be very complex, provided us with a small clue in the form a line at ∼1.0 ppm (¹³C) chemical shift (indicated by a green arrow in the graph). A line at this chemical shift can represent a carbon atom bonded to a silicon atom, implying that the spectral line of interest can correspond to a group of Si-CH type.

Further proof of the involvement of silicon in chemical bonding to organic molecules lies in the detected 1D HSCQ¹H–²⁹Si NMR spectrum (i.e. a spectrum exhibiting signal only for a hydrogen atom in close proximity to and chemically bonded a silicon atom) in green in Fig. 3.6a. This time, the 0.1-ppm line dominates the whole spectrum, or more specifically, is the only discernible feature of the spectrum. The occurrence of this line confirms its assigning as being due to Si-CH groups. The position and narrowness of this line in all the measured spectra suggests that it can be most probably assigned to a (CH₃)₃Si group^∗∗.

Figure 3.7: Diffusion¹H NMR of1sedXDin chloroform, the inset compares the decay of aromatic signal around 7 ppm with a double-exponential fit.

However, the presence of (CH₃)₃Si groups in the Si-coll cannot serve as an ultimate proof of the presence of SiNcs with altered capping. There-fore, a set of ¹H NMR spec-tra for a varying diffusion co-efficient was acquired. As dif-fusion coefficient of the Brow-nian motion in a liquid is a function of the size of the par-ticle, it can, similarly as in DLS, be used to distinguish between particles of different size.

When diffusion NMR spectra^†† were acquired (Fig. 3.7), one could easily see that most of the resonances group around a higher diffusion coefficient (of about 1100µm²/s), while a single strong resonance corresponds to another value of a lower diffusion coeffi-cient (around 158 µm²/s). A simple estimate of the corresponding particles’ diameters according toEq. 3.2^‡‡ yields 0.7 and 5 nm, respectively. Taking into account how rough

k28Si: nuclear spin 0, occurrence 92.2 %,²⁹Si: nuclear spin 1/2 (i.e. can be detected with tradi-tional NMR), occurrence 4.7 %,³⁰Si: nuclear spin 0, occurrence 3.0 %;¹²C: nuclear spin 0, occurrence 99 %,¹³C: nuclear spin 1/2 (i.e. can be detected with traditional NMR), occurrence 1.07 %

∗∗in addition, an acquired HSCQ¹H-¹³C NMR spectrum confirmed this conclusion, however, it is too noisy to be presented

††the type of NMR experiment was double stimulated echo

‡‡this estimate gives the so-called hydrodynamical radius, since it is not completely valid for all situations and in order to determine the particle’s size more precisely it is necessary to use a calibra-tion of diffusion-coefficient/particle-size funccalibra-tion using particles of known sizes and with microscopic behavior in the liquid similar to that of the investigated particles; dynamic viscosity of chloroform of

3.2. Surface chemistry the estimate is, we can assume that the latter size corresponds to single SiNcs in the colloid, while the former is considerably smaller and is likely characteristic of small single molecules.

The resulting correlation of the 0.1-ppm (CH₃)₃Si-group line with a several-nanome-ter-sized particle serves as a convincing argument that CH₃ methyl groups are bonded directly on the surface of single SiNcs (also in accordance withEq. 3.4). The remaining question to be answered is if the SiNcs are covered solely with methyl groups (since e.g. 1D HSCQ ¹H–²⁹Si NMR spectrum in green in Fig. 3.6a seems to consist of this single resonance), or if other passivating agents are present.

The answer to this question is not simple as a result of the already mentioned low signal-to-noise ratio. Nonetheless, it is possible to carry out an analysis of the attenuation of NMR signal for smaller diffusion coefficients, i.e. larger size. For ex-ample, the integral intensity of the band around 7 ppm can be plotted as a function of diffusion coefficient (which is a function of the applied magnetic gradient) as in the inset of Fig. 3.7. When this plot is fitted with expected behavior (approximated with a double-exponential), one can see that the fit diverges from the measured data for smaller values of diffusion coefficient. Consequently, the signal at this band is attenu-ated more slowly than would be expected if this signal were solely due to small, quickly diffusing molecules, and, most likely, larger nanocrystals with lower diffusion coefficient also contribute to the signal in this band.

One of the diffusion ¹H NMR spectra (for low diffusion coefficient, i.e. “large” size, most probably showing only signal corresponding to nanocrystals) is also for comparison plotted together with the other¹H spectra inFig. 3.6a(in red). The comparison clearly illustrates the strong suppression of all the resonances except for the one of the (CH₃)₃Si group. However, several other features besides this resonance still persist (indicated with red arrows). These signals are likely to originate from molecules bonded to SiNcs, which was verified by the quantification of their attenuation characteristics as described above. This result implies that, apart from methyl groups, other molecules with both aromatic and aliphatic groups are bonded on the surface of SiNcs, their identification would require, however, chemically much better-defined mixture eliminating all the

“parasitic” signal from small molecules. The preparation of such a sample, however, is one of the tasks we would like to focus on in the future.