Determining the Surface Structure of Silicated Alumina Catalysts via Isotopic Enrichment and Dynamic Nuclear Polarization Surface-Enhanced NMR Spectroscopy.

Isotopic enrichment of 29Si and DNP-enhanced NMR spectroscopy are combined to determine the detailed surface structure of a silicated alumina catalyst. The significant sensitivity enhancement provided by DNP is vital to the acquisition of multinuclear and multidimensional experiments that provide information on the atomic-level structure of the species present at the surface. Isotopic enrichment not only facilitates spectral acquisition, particularly given the low (1.5 wt %) Si loading, but also enables spectra with higher resolution than those acquired using DNP to be obtained. The unexpected similarity of conventional, CP, and DNP NMR spectra is attributed to the presence of adventitious surface water that forms a sufficiently dense 1H network at the silica surface so as to mediate efficient polarization transfer to all Si species regardless of their chemical nature. Spectra reveal the presence of Si-O-Si linkages at the surface (identified as Q4(3Al)-Q4(3Al)) and confirm that the anchoring of the surface overlayer with the alumina occurs through AlIV and AlV species only. This suggests the presence of Q3/Q4 Si at the surface affects the neighboring Al species, modifying the surface structure and making it less likely AlVI environments are in close spatial proximity. In contrast, Q1/Q2 species, bonded to the surface by fewer covalent bonds, have less of an effect on the surface, and more AlVI species are consequently found nearby. The combination of isotropic enrichment and DNP provides a definitive and fully quantitative description of the Si-modified alumina surface, and we demonstrate that almost one-third of the silicon at the surface is connected to another Si species, even at the low level of coverage used, lowering the propensity for the formation of Brønsted acid sites. This suggests that a variation in the synthetic procedure might be required to obtain a more even coverage for optimum performance. The work here will allow for more rigorous future investigations of structure-function relationships in these complex materials.

Introduction

Silicated aluminas are commonly employed as solid acid catalysts, with applications in a number of processes ranging from ethanol dehydration to hydrocarbon cracking and skeletal isomerization.[1−4] The presence of both Si and Al at the surface generates the mild acidity that is essential to catalytic behavior,[1] but the exact structure of these acidic environments is still debated.[2−4] Early studies of catalytic cracking postulated that Brønsted acidity is attributable to aluminol groups in close proximity to silanols[5] or protons that compensate for the negative charge at the surface.[6,7] More recent investigations (primarily using IR spectroscopy, probe molecule adsorption, and 1H MAS NMR spectra) propose that the catalytic acid sites are bridging Si–OH–Al groups[8−10] or silanols in the vicinity of AlIII, AlIV, or AlV.[2,3] Identifying the true origins of the catalytic response demands an atomic-level description of the reactive surface, but this is far from trivial. The difficulty lies partly in the diverse range of possible surface structures and the typically amorphous character of the materials; for example, the catalytic surface does not display sufficient long-range order to permit structure determination via diffraction-based methods.[11] Vibrational spectroscopy can provide information on the presence of various structural motifs, using molecular probes, but the spectra obtained are often highly complex, leading to difficult and subjective data interpretation. Furthermore, the connectivity between Si and Al is not easily assessed using this approach.[12−14]

While solid-state NMR spectroscopy is ideally suited to investigating the Si–Al connectivity, as it has no requirement for long-range order and is sensitive to small changes in local chemical environments,[15−18] the technique suffers from several practical drawbacks. The inherently low natural abundance of 29Si (4.7%) requires extended acquisition times to obtain spectra with acceptable sensitivity. For many silicated aluminas, this problem is compounded by the low amount of Si, presenting an additional challenge for the implementation of more complex multinuclear and multidimensional experiments.[19] Furthermore, the suitability of conventional 27Al NMR spectroscopy for the characterization of alumina-based catalysts is debatable, because the majority of signals reflect coordination geometries of the bulk rather than the nature of the surface that is responsible for the reaction chemistry.[20]

Recent years have seen a step change in the sensitivity of NMR experiments through the introduction of dynamic nuclear polarization (DNP), where magnetization is transferred from unpaired electrons to nearby nuclei.[21,22] In typical modern DNP experiments for materials, an exogenous solution of a nitroxide biradical in a glass-forming solvent is introduced as a polarization source using incipient wetness impregnation.[23,24] Saturation of the EPR transitions using continuous microwave irradiation polarizes the protons in the solvent, typically via the cross effect.[25,26] Proton spin diffusion distributes this enhanced nuclear polarization across the solvent phase, from where it can be transferred to less receptive nuclei using conventional cross-polarization (CP). This greatly enhances the polarization of surface species relative to the bulk, leading to this approach being referred to as DNP surface-enhanced NMR spectroscopy or DNP-SENS.[23,24] Because of the high signal enhancements available (over 2 orders of magnitude in favorable cases), DNP can overcome the inherently poor sensitivity of many NMR experiments and is being applied increasingly to probe the detailed surface structure of a diverse range of materials.

In recent years, the sensitivity of DNP has provided some new insights into the nature of the interface between Si and Al in catalytically important amorphous silicated aluminas.[27−30] This has included the observation (using 29Si NMR) of isolated SiOH species on the surface,[27,28] the presence of Brønsted acid sites (and the determination of the O–H bond lengths),[29] and some insight into the Al/Si connectivity using two-dimensional correlation experiments.[30] However, the determination of true structure–function relationships requires a fully quantitative description of surface structure, which has not yet been able to be demonstrated, and is the focus of the present work. The DNP enhancement factor, ε, is determined from the ratio of signal intensities with and without microwave irradiation, and is dependent on temperature, microwave power, concentration of the exogenous radical source, and the extent of surface wetting.[21,22] However, proton density also has a significant effect on the absolute amount of signal observed, a direct consequence of the complex pathway of polarization transfer from the electron to the target nucleus. Given the expected dependence of the 1H–29Si CP transfer efficiency on the local proton density, it would be expected that signal intensities would be highly dependent on the proximity of a nucleus to hydroxyl groups at the sample surface.[24] Resonances from lower-order Q species (where Q denotes a species of structure Si(OT)(OH)4–), which possess a higher local proton density, are expected to be amplified to a greater extent than higher-order Q analogues. This potential variation in signal enhancement should, therefore, limit DNP (and similarly CP) measurements to a qualitative, or at best semiquantitative, description of surface structure, which in turn places limitations on the determination of accurate structure–function relationships. Such effects of proton proximity are clearly demonstrated by the work of Lelli et al., who investigated phenol functionalized silica surfaces.[24] At very short contact times, only 29Si centers in close proximity to 1H were enhanced. Signal intensity associated with Q sites that are further removed from 1H increased with longer CP mixing times as polarization was transferred from more distant spins.

While the expected variation in CP and DNP signal intensity with local proton density may limit these approaches to providing a more qualitative description of surface structure, it would, nonetheless, offer important and detailed information on the atomic-scale environment, which can be vital for spectral assignment. Ideally, therefore, spectra edited on the basis of spatial proximity would be compared to more quantitative spectra (i.e., using conventional experiments that do not rely on any polarization transfer and have sufficiently long recycle intervals to account for any differences in relative relaxation). The obvious solution here is to exploit isotopic enrichment, improving the sensitivity of the conventional NMR experiment to such an extent that spectra can be acquired on a reasonable time scale, and spectral line shapes can then be compared to those acquired using CP/DNP. Enrichment also has additional advantages for the sensitivity of heteronuclear, and particularly homonuclear, two-dimensional correlation NMR experiments, made possible using DNP but ultimately limited by the low abundance of NMR-active Si species.

In this work, we exploit both isotopic enrichment of 29Si and DNP enhancement to determine the detailed surface structure of a Si-γ-Al2O3 material with 1.5 wt % Si doping. This combined strategy enables us not only to obtain the quantitative NMR spectra that are so key to understanding surface structure, but also to exploit the significant sensitivity advantages offered by DNP and acquire multidimensional experiments that simply would not be possible otherwise. The combination of these two approaches enables a confident spectral assignment and the determination of the type and, more importantly the proportion of surface species present. The understanding we have gained here will be used to direct future synthetic approaches for surface modification of similar materials, confident that subsequent quantitative analysis will be possible using the approaches introduced in this work, and insight into the structure–function relationships for alumina-based catalysts can then be unveiled.

Experimental Details

Synthesis

Si-γ-Al2O3 (1.5 wt % Si) materials were synthesized by a conventional wet impregnation procedure. γ-Al2O3 (Sasol, 98%) was impregnated with either conventional (Sigma-Aldrich, 98%) or 99% 29Si-enriched (Cortecnet, >95%) tetraethyl orthosilicate dissolved in dry ethanol, in an inert (Ar) atmosphere. Samples were then dried at 60 °C in vacuo and subsequently calcined in air at 520 °C for 2 h (ramp rate of 10 °C min–1). The final composition was confirmed by elemental analysis (ICP OES, see the Supporting Information). The initial alumina contained spherical particles with an average diameter of 75 μm. The silicated aluminas were characterized using N2 adsorption measurements (see the Supporting Information). Dehydrated 29Si-enriched Si-γ-Al2O3 was dried in vacuo (150 °C, ∼12 h) and subsequently packed into a 4 mm ZrO2 rotor in an inert (N2) atmosphere.

NMR Spectroscopy

NMR spectra were acquired using a Bruker Avance III spectrometer, equipped with a 9.4 T widebore magnet operating at Larmor frequencies of 400.13 MHz for 1H, 104.3 MHz for 27Al, and 79.46 MHz for 29Si. Powdered samples were packed into 4 mm ZrO2 rotors and rotated at magic angle spinning (MAS) rates of 10–14 kHz, using a conventional 4 mm HX probe. Spectra were acquired using radio frequency field strengths of ∼90 kHz for 1H, ∼110 kHz for 27Al, and ∼80 kHz for 29Si. Typical recycle intervals were 1 s for 1H, 3 s for 27Al, and 120 s for 29Si. (For 29Si NMR, T1 values were estimated to be much longer than 120 s, as discussed in a later section, but little difference in relative relaxation was observed.) 1H spectra were acquired using the DEPTH pulse sequence for probe background suppression.[31]29Si spectra were acquired using either single pulse excitation (DP), DEPTH pulse sequences, or cross-polarization. For CP spectra, transverse magnetization was obtained from 1H using contact pulse durations of between 0.1 and 10 ms (ramped 90–100% for 1H) and a recycle interval of 1 s. TPPM 1H decoupling (∼90 kHz)[32] was applied during acquisition. For 1H–29Si CP HETCOR NMR spectra, contact pulse durations of 0.5 and 3 ms were used. Sign discrimination in the indirect dimension was achieved using the quadrature detection method of Marion et al.[33]27Al spectra were obtained using either single pulse excitation or CP. For CP spectra, transverse magnetization was obtained from 1H using a contact pulse duration of 0.8 ms (ramped for 1H) and a recycle interval of 1 s. Chemical shifts are shown (quoted in ppm) relative to (CH3)4Si for 1H and 29Si and 1 M Al(NO3)3 (aq) for 27Al, measured using secondary references of l-alanine for 1H (NH3 δ = 8.5 ppm), octakis(trimethylsiloxy)silsesquioxane (Q8M8) for 29Si (OSi(CH3)3 δ = 11.5 ppm), and aluminum acetylacetonate for 27Al (δiso = 0.0 ppm, CQ = 3.0 MHz, ηQ = 0.16).

DNP NMR experiments were performed using a Bruker Avance I spectrometer, equipped with a 9.4 T widebore magnet operating at Larmor frequencies of 400.21 MHz for 1H, 104.29 MHz for 27Al, and 79.50 MHz for 29Si. A 9.7 T gyrotron magnet was utilized for the generation of microwaves, operating at a frequency of 263 GHz. The field of the main (9.4 T) magnet was set such that microwave irradiation occurred at the 1H positive enhancement maximum of nitroxide biradicals. Incipient wetness impregnation of powdered samples (∼25 mg) was performed with a solution (16–24 μL) of the nitroxide biradical polarizing agent TEKPol in 1,1,2,2-tetrachloroethane (TCE) (14–16 mM).[34,35] Impregnated samples were packed into 3.2 mm sapphire or ZrO2 rotors, and frozen at 100 K inside a 3.2 mm low-temperature MAS probe using dry N2 as the bearing and drive gas. Samples were typically subjected to multiple thawing cycles by ejecting the sample into the catcher in the room-temperature region of the probe, to minimize the amount of oxygen in the biradical solution (which would decrease DNP enhancements).[36] Samples were rotated at MAS rates between 8 and 12.5 kHz. Ramped (90–100 or 80–100%) CP was used to transfer polarization from 1H to 29Si or 27Al. SPINAL 1H decoupling[37] was applied during acquisition. For 1H–27Al CP experiments, a low-power 27Al radiofrequency field was used to ensure efficient spin locking of the quadrupolar nucleus.[38,39] Typical DNP enhancements (calculated by comparing spectra acquired with and without microwave irradiation) were ε = 92 (29Si) and ε = 112 (27Al). Two-dimensional (2D) 29Si–29Si double-quantum correlation spectra were acquired using a refocused INADEQUATE experiment[40] using τJ intervals of between 3.2 and 16 ms. 2D 29Si–27Al scalar (through-bond) correlation spectra were acquired with a refocused J-INEPT experiment[41] and a τJ interval of 6 ms. 2D 29Si–27Al dipolar (through-space) correlation spectra were acquired with a dipolar refocused D-INEPT experiment,[42] using REDOR[43] for heteronuclear 29Si–27Al dipolar recoupling. In all cases, initial 29Si polarization was generated via 1H–29Si CP with contact pulse durations of between 3 and 3.5 ms. For all 2D experiments, the quadrature detection method of States et al.[44] was used to achieve sign discrimination in the indirect dimension. Chemical shifts are shown (quoted in ppm) relative to (CH3)4Si (1H and 29Si) and 1.0 M Al(NO3)3 (aq) (27Al), measured using an internal reference of TCE (1H δ = 6.2 ppm). Line shape fitting was carried out using dmfit.[45]

Results and Discussion

Figure shows 29Si MAS NMR spectra of 99% 29Si enriched Si-γ-Al2O3 (1.5 wt % Si), acquired using direct polarization (DP), CP, and DNP. All DP and CP experiments were carried out at room temperature (298 K) on samples with no radical added, while DNP experiments were carried out at low temperature (100 K). Given the inherently nonquantitative nature of CP and the additional surface sensitivity of DNP, the spectral line shapes are remarkably similar (as shown in Figure S1, where the line shapes are overlaid). The DP and CP spectra exhibit better resolution than the DNP spectrum (although the DP spectrum has poorer sensitivity), as a result either of increased relaxation arising from the presence of the radical or, more likely, of the lower temperature of the experiments (and reduced molecular motion), leading to a broader distribution of shifts. Despite the much greater sensitivity of the DNP spectrum, it is not possible to decompose the line shape unambiguously into individual components. However, four distinct environments are discernible in the DP and CP NMR spectra: three sharp components at high frequency and a lower intensity, broader resonance at lower frequency. It is worth noting that the DP spectrum would have taken ∼311 days to acquire at natural abundance (see Figure S2). Figure shows the variation in the spectrum as a function of the CP contact time, τCP. In contrast to the materials studied in ref (24), there are only very small differences in the spectral line shape as τCP increases. It is not clear if this result suggests that no Q4 species are present (i.e., all Si are connected to at least 1 OH group), certainly possible at the low weight loading considered here.[46,47] It is difficult to assign species in this spectrum simply on the basis of δ, as the expected −10 ppm change with the number of coordinated bridging oxygen species is complicated in aluminosilicates by an additional shift to higher frequency of 5–8 ppm per next nearest neighbor (NNN) Al.[48,49] Therefore, it is not possible to unambiguously confirm the presence, or absence, of Q4 species in the spectra of Si-γ-Al2O3, although the signal observed extends over the region expected for these species.

Figure 1

29Si and 1H–29Si (9.4 T, 10–14 kHz MAS) NMR spectra of 99% 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si), acquired using (a) direct polarization (DP), (b) cross-polarization (CP), and (c) DNP. DP and CP spectra were acquired at room temperature on samples that had no radical added. The DNP spectrum was performed at 100 K. Spectra are the result of averaging (a) 504, (b) 14 400, and (c) 32 transients with recycle intervals of (a) 120, (b) 1, and (c) 3 s. For CP and DNP spectra, a contact pulse duration of 3 ms was used.

Figure 2

(a) 1H–29Si (9.4 T, 10 kHz MAS) CP NMR spectra of 99% 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si), acquired by averaging 14 400 transients separated by a recycle interval of 1 s, using τCP values between 0.1 and 10 ms. (b) Comparison of 1H–29Si (9.4 T, 10 kHz MAS) CP NMR spectra of hydrated (black line) and dehydrated (red line) 99% 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si). Spectra are the result of averaging 14 400 (hydrated) and 34 000 (dehydrated) transients separated by a recycle interval of 1 s. Polarization transfer (from 1H) was achieved using τCP values of 3 ms. In each case, the spectral intensities have been normalized.

The features in the DP and CP spectral line shapes in Figure do suggest that different Q species are present, making it all the more surprising that little variation in the line shape is observed either between the two experiments or as a function of the CP contact time. It is clear from Figure that the polarization transfer in CP remains equally efficient for all species, irrespective of the number of hydroxyl groups attached. However, it is possible that the presence of adventitious surface water, known to form extensive and strong H-bonding with silanols,[50,51] could affect the spectral intensities observed. To determine if this adventitious water is playing a role in the CP dynamics, the 29Si enriched Si-γ-Al2O3 was dehydrated in vacuo at 150 °C and packed into a ZrO2 rotor in a glovebox. Relatively mild conditions were used for dehydration to avoid any surface dehydroxylation. Verification of dehydration was obtained using 1H MAS NMR, as shown in Figure S3. Figure b compares the 29Si CP MAS spectra of hydrated and dehydrated Si-γ-Al2O3 and reveals a change in the relative intensities of the spectral components, with an increase in signal intensity of the peak at δ = −78 ppm, confirming the higher density of OH groups. The DP and CP spectra of dehydrated Si-γ-Al2O3 are less similar, as shown in Figure S4, with a relative increase in the intensity of the signal at higher δ in the CP spectrum. Crucially, the DP NMR spectrum remains largely unaffected by dehydration, an indication that surface structure has remained intact following the high temperature treatment (Figure S5). From these observations, it is clear in this case that the H-bonded water forms a sufficiently dense 1H network at the silica surface as to mediate efficient polarization transfer to all Si species regardless of their chemical nature and OH functionality.

Figure a compares 1H–29Si CP HETCOR NMR spectra of hydrated and dehydrated Si-γ-Al2O3 materials, and shows appreciable differences in the extent of correlation between 29Si and 1H upon the removal of water. When dehydrated, higher-order Q species no longer correlate with surface protons because, in the absence of surface water, the 29Si spectrum is influenced more significantly by local proton (hydroxyl) density. It is clear from Figure b that dehydration also results in a more significant variation in the spectral line shape with contact time. At sufficiently long contact times, correlations to all Si species are observed, but intensity is lost from the region between −85 and −100 ppm as contact time is reduced. Therefore, we conclude that Q4 species are, most likely, present at the silicated surface and, importantly, that the extent of signal amplification remains constant, irrespective of local hydroxyl density, by virtue of adventitious adsorbed water when the sample is hydrated (or stored under ambient conditions). Thus, perhaps surprisingly, as long as the surface remains sufficiently hydrated, as is the case here, DNP NMR spectra of such materials may be interpreted quantitatively, and an accurate description of structure–function relationships can be obtained.

Figure 3

1H–29Si (9.4 T, 10 kHz MAS) CP HETCOR NMR spectra of 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si), demonstrating the effect of (a) the hydration level of the sample and (b) τCP on the dehydrated material. In (a), the spectrum of the hydrated material is the result of averaging 880 transients separated by a recycle interval of 1 s, for each of 20 t1 increments of 100 μs. The spectrum of the dehydrated material is the result of averaging 1600 transients separated by a recycle interval of 1 s, for each of 18 t1 increments of 50 μs. A CP contact time of 0.5 ms was employed in both instances. In (b), both spectra were acquired by averaging 1600 transients separated by a recycle interval of 1 s, for each of 18 t1 increments of 50 μs.

The combination of 29Si isotopic enrichment and DNP NMR spectroscopy results in a significant signal enhancement that provides access to two-dimensional experiments that may otherwise require prohibitively long acquisition times. In particular, correlations exploiting through-bond J couplings can be valuable sources of information on the nature of the surface structure. Figure a shows a 29Si CP MAS INADEQUATE[40] DNP SENS spectrum of hydrated Si-γ-Al2O3 (1.5 wt % Si). Despite the low Si content, the use of isotopic enrichment combined with DNP permits spectral acquisition in only 4 h. Signal is observed between −80 and −100 ppm, suggesting that only higher-order Q species are connected to Si; that is, the three sharper peaks at more positive shift result from isolated Q(nAl) species. Signal is observed over a range of ∼20 ppm in δ1, possibly indicating that this results from more than one chemical species, for example, Q3 and Q4 species. However, the correlation peak lies primarily along the δ1 = 2δ2 diagonal in the two-dimensional spectrum, confirming that Si species are only covalently connected to those with very similar shift and, hence, very similar environments. This would suggest the signal probably arises from only Q4/Q4 or Q3/Q3 correlations (rather than Q4/Q3, for example), and the loss of signal at −90 ppm at longer τCP times in the 1H–29Si HETCOR spectrum in Figure b supports a more likely assignment of Q4/Q4 species for this peak. Spectra acquired with a variety of J evolution times (shown in Figure S6) also show only autocorrelation signal. Projections of the spectra onto the δ2 axis (shown in Figure b) show that signal shifts to higher δ as τJ increases, indicating a positive correlation between 29Si nuclear shielding and the homonuclear J coupling. Such a correlation has also been observed for other (alumino)silicate materials, and was related to changes in the Si–O–Si bond angle using ab initio cluster calculations.[52,53] While it was shown that the exact values of δ and J vary with the cations present (and so cannot be directly related to the materials studied here), it seems likely that the same structural change is likely to be responsible, thus suggesting that the Si–O–Si bond angle is decreasing (by ∼7–8°) between the signals seen at −93 ppm (when τJ = 3.2 ms) and −85 ppm (when τJ = 16 ms).

Figure 4

29Si (9.4 T, 12.5 kHz MAS) refocused CP INADEQUATE DNP NMR spectrum of hydrated 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si). The spectrum is the result of averaging 96 transients separated by a recycle interval of 3 s for each of the 48 t1 increments of 80 μs, with a τJ value of 9.6 ms. Polarization transfer was achieved using a τCP value of 3 ms. (b) Overlay of (δ2) projections of two-dimensional refocused INADEQUATE DNP NMR spectra (shown in the Supporting Information) as a function of τJ time. The 1H–29Si CP DNP NMR spectrum is also shown for comparison.

To understand the interfacial chemistry between the silica surface overlayer and the γ-Al2O3 structure upon which it resides, heteronuclear correlation experiments (29Si–27Al refocused INEPT41) were performed, again exploiting DNP to improve sensitivity. This experiment can probe scalar (through-bond) connectivity, or can be adapted in the solid state to actively recouple the dipolar interaction and provide information on through-space proximities. Figure a shows a 29Si–27Al dipolar INEPT DNP NMR spectrum of hydrated Si-γ-Al2O3, acquired using REDOR to recouple the dipolar interaction.[43] This reveals that Si is close in space to four-, five-, and six-coordinate Al (i.e., AlIV, AlV, and AlVI). As Si is present only as a surface overlayer, the spectrum contains only Al species that are close to the surface. As shown in the Supporting Information, the 27Al spectrum of bulk γ-Al2O3 contains signals that can be attributed to AlIV and AlVI species only.[54,55] However, the surface of γ-Al2O3 has been shown to contain AlV species. These can be seen using CP, where magnetization is transferred from surface-based 1H species, resulting in additional signal at δ ≈ 35 ppm (see the Supporting Information). DNP NMR spectra have also demonstrated the presence of AlV at the surface of γ-Al2O3.[56] Using a filtration experiment, where signals close to the surface dephase due to their stronger dipolar couplings to 1H, Lee et al. demonstrated that AlV resides only in the first surface layer. The 27Al MAS NMR spectrum of Si-γ-Al2O3, shown in the Supporting Information, reveals a similar picture, with resonances corresponding to AlIV and AlVI in the bulk material, while the CP spectrum also shows the presence of AlV at the surface, in agreement with recent work probing the nature of Brønsted acid sites.[57] Although AlV species are found at the surface of unmodified γ-Al2O3, it has been suggested that additional AlV is formed at the interface between the Si and alumina, with the proportion of these varying with the Si content. Through-bond connectivity can be probed using the 29Si–27Al INEPT DNP NMR spectrum acquired without dipolar recoupling (Figure b). In contrast to the spectrum in Figure a, this reveals that Si is covalently connected via AlIV and AlV anchor points. Thus, only a spatial proximity to AlVI exists. It is also interesting to note from Figure a that AlIV/V species show the strongest (through-space) correlation with Si signal near −84 ppm, while AlVI species are more strongly correlated with signal at higher δ (i.e., lower-order Q species), suggesting these are more frequently found in close proximity at the surface.

Figure 5

29Si–27Al (9.4 T, 10 kHz MAS) CP INEPT DNP NMR spectrum of hydrated 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si), with transfer via the (a) dipolar and (b) scalar coupling. Spectra are the result of averaging (a) 48 and (b) 224 transients separated by a recycle interval of 2 s for each of 32 t1 increments of 100 μs. In (b), recoupling of the dipolar interaction was carried out using 4 REDOR blocks of 8 rotor cycles (of 100 μs) in duration. Polarization transfer was achieved using a τCP value of 3 ms.

Extracting quantitative information from the broadened 29Si resonance observed using DNP NMR is nontrivial. The conventional 29Si MAS NMR spectrum of the 29Si-enriched material exhibits better resolution, and, when combined with the information obtained from CP spectra, and particularly from the projection of the single-quantum dimension of the INADEQUATE spectra, a more robust deconvolution is possible, as shown in Figure , with parameters given in Table . Sharp components can be identified at −77, −80, and −83 ppm (where constraints on the positions and line shapes were determined using variable contact time CP experiments), with a broader component centered at ca. −89 ppm. The position and line shape of the latter are determined from the INADEQUATE spectra. From previous literature,[15,48,49] the three sharp signals can be assigned as Q1(1Al) (resonance 1), Q2(2Al) (resonance 2), and Q3(3Al)/Q4(4Al) (resonance 3) species. The opposing shifts that are induced by increased condensation of Si–O tetrahedra and substitution of Si with Al lead to an inevitable overlap of resonance frequencies for some species. For this reason, it is difficult to discriminate between Q3(3Al) and Q4(4Al) on the basis of chemical shift alone, and it is possible that both species contribute to the signal at −83 ppm. However, the presence of signal at −83 ppm in the 1H–29Si HETCOR spectrum in Figure b at short contact times suggests a significant contribution to the intensity at this point must arise from Q3(3Al) Si centers. The projection of the 29Si DNP NMR INADEQUATE spectrum in Figure b confirms that the component at −89 ppm contains primarily Si species within Si–O–Si linkages. As discussed above, the observation of signal along the autocorrelation diagonal, and the loss of this signal in the 1H–29Si HETCOR spectrum, suggests it results principally from interconnected Q4(3Al) species, that is, (OAl)3Si–O–Si(OAl)3 linkages at this loading. At first sight, it is not perhaps clear why most Q4 Si species are linked to a second Si center. The low loading of Si in these samples ensures that most Si species are bonded only to Al (i.e., Q1(1Al), Q2(2Al), and Q3(3Al)). As Si forms a surface layer on the alumina, it is less likely that it embeds to make four bonds to the surface Al, and that Q4 Si species form (primarily, but perhaps not exclusively) when they are able to bond via a bridging oxygen to a second Si species on the surface.[57−59] As shown in Table , from the spectrum in Figure we find Q1(1Al) ≈ 22%, Q2(2Al) ≈ 14%, Q3(3Al)/Q4(4Al) ≈ 34%, and Q4(3Al) ≈ 30%, suggesting that almost one-third of the silicon at the surface is connected to another Si species, even at the low level of coverage used. As the Si–O–Si connectivity increases, the propensity for the formation of Brønsted acid sites will diminish. Thus, a variation in the synthetic procedure might be required to obtain a more even coverage for optimum performance. It should be noted that, although the recycle interval used for the 29Si DP MAS spectrum was 120 s, the T1 value was estimated later to be on the order of ∼1.5 h, making acquisition of a truly quantitative spectrum practically unfeasible. Although this may result in some uncertainty in the exact proportion of each species present, little difference was observed in the relative relaxation of the different Si species at shorter recycle intervals. Deconvolution of the 29Si spectrum and assignment of the contributions of the component resonances would have been almost impossible using conventional NMR spectroscopy, and difficult by either DNP (due to the lower resolution) or isotopic enrichment (due to the lower sensitivity) alone.

Figure 6

(a) 29Si (9.4 T, 14 kHz MAS) experimental (blue) and simulated (red) NMR spectra of hydrated 29Si-enriched Si-γ-Al2O3 (1.5 wt % Si). Also shown are the individual components of the fit (green). The experimental spectrum is the result of averaging 504 transients separated by a recycle interval of 120 s. (b) Assignment of the proposed structural motifs present.

Table 1

29Si Chemical Shifts (δiso), Relative Intensities, and Assignments for the Contributions to the 29Si MAS NMR Spectrum of 99% 29Si-Enriched Si-γ-Al2O3 (1.5 wt % Si) Shown in Figure

component	δiso (ppm)	relative intensity (%)	assignment
1	–77 (1)	22 (2)	Q1(1Al)
2	–80 (1)	14 (3)	Q2(2Al)
3	–83 (1)	34 (2)	Q3(3Al)/Q4(4Al)
4	–89 (2)	30 (2)	Q4(3Al)

Conclusions

We have exploited a combination of isotopic enrichment and DNP to provide a definitive and fully quantitative description of the surface structure of Si-modified alumina catalysts. Comparison of DP, CP, and DNP 29Si NMR spectra surprisingly reveals very similar line shapes, demonstrating that 1H–29Si CP transfer efficiency, and the extent of signal enhancement, is constant and independent of the proximity of a nucleus to surface hydroxyl groups. We have attributed this unexpected behavior to the presence of adventitious surface water in these highly hygroscopic materials. This H-bonded water forms a sufficiently dense 1H network at the silica surface as to mediate efficient polarization transfer to all Si species regardless of their chemical nature and OH functionality. Upon dehydration, this network is disrupted, and the transfer efficiency becomes more dependent on the chemical nature of the species present. This leads to the unforeseen conclusion that, if sufficiently hydrated, CP (and DNP) NMR spectra of the Si-modified alumina surface can be interpreted essentially quantitatively, allowing for accurate and detailed determination of structure–function relationships.

Despite the significant sensitivity advantage afforded by DNP, spectra exhibit comparatively lower resolution, most likely as a result of the lower temperature at which experiments are performed. To obtain an accurate deconvolution of the spectral line shapes, and to determine the relative proportion of each species present, isotopic enrichment and the acquisition of non DNP-enhanced spectra at the low Si loading present are vital. However, the combination of DNP and isotopic enrichment provides access to two-dimensional experiments that would otherwise require prohibitively long acquisition times given the low Si content. 29Si INADEQUATE experiments facilitate the identification of Si–O–Si units in the Si-γ-Al2O3 structure as interconnected Q4(3Al) species. Heteronuclear 29Si–27Al INEPT experiments confirm that Si is present only as a surface overlayer and reveal that the reaction of TEOS with γ-Al2O3 occurs via condensation reactions at AlIV and AlV anchoring points. Furthermore, the corresponding dipolar INEPT spectrum suggests that the presence of Q3/Q4 Si at the surface affects the neighboring Al species, modifying the surface structure and making it less likely AlVI environments are in close spatial proximity. In contrast, the presence of Q1/Q2 species, bonded to the surface by fewer covalent bonds, has less effect, and more AlVI species are then found nearby.

The combination of the increased amount of information available as a result of the DNP enhancement, and the ability to obtain quantitative spectra using isotopic enrichment, affords a more rigorous quantitative interpretation of 29Si spectra and a detailed understanding of the nature of the Si–alumina interface. The ability to accurately describe surface structure will allow for more rigorous investigations of structure–function relationships and the future design of synthetic protocols that permit a tailoring of surface sites and, ultimately, catalytic performance.