The Surface of Nanoparticle Silicon as Studied by Solid-State NMR

The surface structure and adjacent interior of commercially available silicon nanopowder (np-Si) was studied using multinuclear, solid-state NMR spectroscopy. The results are consistent with an overall picture in which the bulk of the np-Si interior consists of highly ordered (“crystalline”) silicon atoms, each bound tetrahedrally to four other silicon atoms. From a combination of 1H, 29Si and 2H magic-angle-spinning (MAS) NMR results and quantum mechanical 29Si chemical shift calculations, silicon atoms on the surface of “as-received” np-Si were found to exist in a variety of chemical structures, with apparent populations in the order (a) (Si–O–)3Si–H > (b) (Si–O–)3SiOH > (c) (HO–)nSi(Si)m(–OSi)4−m−n ≈ (d) (Si–O–)2Si(H)OH > (e) (Si–O–)2Si(–OH)2 > (f) (Si–O–)4Si, where Si stands for a surface silicon atom and Si represents another silicon atom that is attached to Si by either a Si–Si bond or a Si–O–Si linkage. The relative populations of each of these structures can be modified by chemical treatment, including with O2 gas at elevated temperature. A deliberately oxidized sample displays an increased population of (Si–O–)3Si–H, as well as (Si–O–)3SiOH sites. Considerable heterogeneity of some surface structures was observed. A combination of 1H and 2H MAS experiments provide evidence for a substantial population of silanol (Si–OH) moieties, some of which are not readily H-exchangeable, along with the dominant Si–H sites, on the surface of “as-received” np-Si; the silanol moieties are enhanced by deliberate oxidation. An extension of the DEPTH background suppression method is also demonstrated that permits measurement of the T2 relaxation parameter simultaneously with background suppression.

. 50 MHz 13 C CP-MAS (3.5 kHz) spectra of (A) np-Si and (B) 1% HMB mixed with silica gel. In both cases 15,000 transients were collected, using 1 s repetition delays and a separately determined background signal was subtracted. In both samples the same probe and rotor configuration was used. Table 1) Results shown here are presented in three parts-the experimental spectrum, the simulated spectrum based on the deconvolved spectral contributions and the difference between the first two. Figure S2 shows the experimental and simulated proton-decoupled 29 Si CP-MAS NMR spectra of np-Si evacuated at 150 °C. The experimental spectrum, as with all but that of Figure S3, was taken with a CP contact time (CT) of 14 ms.  The simulated spectrum of Figure S2 consists of six heavily overlapping signals, each with a unique chemical shift, whose spectral characteristics are summarized, along with the other cases in this Section 2, in Table 1 and Table S1. One signal centered at −14 ppm has a Gaussian line shape, while the five others at −74, −84, −91, −100 and −109 ppm have Lorentzian line shapes. As with all of the spectra in this section, the use of a mixed (linear combination of Lorentzian and Gaussian) line shape for any of the signals did not improve perceptibly the match between the experimental and simulated spectra. The signal at −14 ppm differs from the rest in having a significantly larger line width of 28 ppm, compared with 9 to 12 ppm for the others. No spinning sidebands are observed for any signal. Figure S3 shows the experimental and simulated proton-decoupled 29 Si CP-MAS spectra of np-Si evacuated at 150 °C, taken with a 1.0 ms CP contact time. The line shape and line width of each signal in Figure S3 are very similar to the values of the corresponding signals in Figure S2, with the exception of the signal at −74 ppm, where the short CP contact time result is several ppm narrower than the long CP contact result. Although the difference is significant, it is small and its significance uncertain. Figure S4 shows the experimental and simulated proton-coupled 29 Si CP-MAS NMR spectra of np-Si evacuated at 150 °C. The experimental spectrum was recorded without proton decoupling and using the same long CP contact time as used in Figure S2 Figure S5 shows the experimental and simulated proton-decoupled 29 Si CP-MAS NMR spectra of 2 H 2 O-treated np-Si, obtained using a long CP contact time (14 ms). The simulated spectrum of Figure S5 consists of five heavily overlapping signals ( Table 1). The signal at −14 ppm exhibits a Gaussian line shape. Signals at −74, −83, −89, −99 and −108 ppm have Lorentzian line shapes. Figure S6 shows the experimental and simulated proton-decoupled 29 Si CP-MAS NMR spectra of oxidized np-Si, obtained using a long CP contact time (14 ms). The simulated spectrum of Figure S6 consists of three heavily overlapping signals ( Table 1) Figure S7 shows the experimental and simulated proton-decoupled 29 Si CP-MAS NMR spectra of oxidized-then-1 H 2 O-treated np-Si, obtained using a long CP contact time (14 ms). The simulated spectrum of Figure S7 consists of three heavily overlapping signals (Table 1). Signals at −89, −99 and −109 ppm have Lorentzian line shapes. Figure S8 shows the experimental and simulated proton-decoupled 29 Si CP-MAS NMR spectra of np-Si that has been treated with 1 H 2 O, obtained using a long CP contact time (14 ms). The simulated spectrum of Figure S8 consists of six heavily overlapping signals ( Table 1)

DEPTH-Echo Method
There are occasions when, in the course of measuring the NMR spectrum of a substance, especially one yielding relatively weak signals, the spectrum is contaminated by signals arising from the materials used to construct the NMR probe. Of course, NMR probes are generally constructed in such a fashion to avoid this spectral contamination but there are circumstances where it is unavoidable and the spectroscopist must adopt measures to eliminate it. Several strategies have been developed over the years to accomplish this decontamination; each has strengths and weaknesses, which makes it more or less applicable under specific circumstances.
One strategy is to record two NMR spectra, one with and one without the substance under study (the sample) in the probe (all other experimental variables are held constant) and to take the algebraic difference of these two spectra. This difference corresponds to the NMR spectrum of the sample free of contamination. The process is often referred to as "background suppression through subtraction", where the term background refers, in particular, to the probe materials' signals and more generally to all signals other than those arising from the sample. It will be recognized that, if this prescription is followed exactly, the signal-to-noise ratio of the decontaminated spectrum suffers by a factor of √2 and that the total NMR instrument time is doubled over that needed to obtain the original NMR spectrum with the original S/N. The first drawback can be avoided by the introduction of an additional step in the process where the so-called "background spectrum", that is, the one without the sample, is replaced by a simulated version of it prepared by the spectroscopist and designed to match all of the observable spectral features but without the instrumental noise. The second drawback can be mitigated but perhaps not totally eliminated by characterizing the probe's background spectrum using a smaller amount of signal averaging or fixing upon the experimental conditions so that the background spectrum can be measured once and applied to many other measurements.
Other methods rely on the difference(s) between certain spectroscopic parameters of the substance under study and the probe materials, such as the longitudinal relaxation time, T 1 , or the transverse relaxation time, T 2 . Methods such as these can produce decontaminated spectra-provided there is a set of conditions where the spectroscopic distinction between the substance under study and the background signals is clear.
Another and powerful method, known as DEPTH [1,2], relies on a predominant RF design characteristic of NMR probes-the magnitude of the magnetic field rotating at the spectrometer's carrier frequency, B 1 , is spatially inhomogeneous. More precisely, the sequence capitalizes on the fact that there is a substantial difference between the magnitude of B 1 at the sample and at background positions. This difference, together with the behavior of nuclear spins under the RF field-induced rotations, can be used to preserve the sample's signals and to suppress the background's signals ( Figure S10). Figure S10. Comparison of the structure of pulse sequences for Bloch Decay, DEPTH-1 (with one π pulse) and DEPTH-2 (with two π pulses). The pattern of RF phases in the rotating frame used to coherently average data varies among the pulse sequences.
The principles of the mechanism of background suppression using DEPTH-style pulse sequences have been discussed extensively [1]. In the present work we deliberately include systematic variation of several free precession times in the DEPTH pulse sequences. See Figure S11. These varying times have been added for the purpose of suppressing the proton background signals of the NMR probe arising from the proton-containing materials used in its construction. These timing parameters also offer an additional degree of experimental flexibility to tease apart the heavily overlapping manifold of proton signals arising from the samples by taking advantage of differing T 2 values. A series of separate experiments using crystalline solid-state samples (data not presented) demonstrated that the combination of DEPTH refocusing pulses and the inter-pulse times do not interfere with the suppression of background signals nor do they perturb the values of T 2 relaxation times measured on these samples by traditional methods. Figure S11. Comparison of the structure of pulse sequences for chemical shift echo, DEPTH (with two π pulses) and DEPTH-echo (with two π pulses). The patterns of RF phases in the rotating frame for DEPTH2 and DEPTH-echo are unaffected by the inter-pulse spacing and are identical for the two sequences. For each experiment, the experimental (noisy lines) and corresponding simulated spectra (smooth lines) for each value of τ total time are overlapped to permit comparison. The derived intensities and widths of the spectral contributions are summarized in Table 2 and Table S2.

Simulation/Deconvolution of DEPTH-Echo 1 H MAS Spectra
All signals exhibit Lorentzian line shapes. As with all of the spectra in this section, the use of a mixed (linear combination of Lorentzian and Gaussian) line shape for any of the signals did not improve the match between the experimental and simulated spectra.          Figure S21 contains experimental and simulated dipolar-dephasing 29 Si NMR CP-MAS spectra of np-Si samples as described in the main text. In the figure, the experimental (noisy lines) and corresponding simulated (smooth lines) spectra for each value of dipolar dephasing time are overlapped to permit comparison between them.

Simulation/Deconvolution of 29 Si CP-MAS Dipolar-Dephasing Results
One signal centered at −14 ppm has a Gaussian line shape, while the five others at −74, −84, −91, −100 and −109 ppm have Lorentzian line shapes. As with all of the spectra in this section, the use of a mixed (linear combination of Lorentzian and Gaussian) line shape for any of the signals did not improve the match between the experimental and simulated spectra.