High-Field One-Dimensional and Two-Dimensional 27Al Magic-Angle Spinning Nuclear Magnetic Resonance Study of θ-, δ-, and γ-Al2O3 Dominated Aluminum Oxides: Toward Understanding the Al Sites in γ-Al2O3.

Herein, a detailed analysis was carried out using high-field (19.9 T) 27Al magic-angle spinning (MAS) nuclear magnetic resonance (NMR) on three specially prepared aluminum oxide samples where the γ-, δ-, and θ-Al2O3 phases are dominantly expressed through careful control of the synthesis conditions. Specifically, two-dimensional (2D) multiquantum (MQ) MAS 27Al was used to obtain high spectral resolution, which provided a guide for analyzing quantitative 1D 27Al NMR spectra. Six aluminum sites were resolved in the 2D MQ MAS NMR spectra, and seven aluminum sites were required to fit the 1D spectra. A set of octahedral and tetrahedral peaks with well-defined quadrupolar line shapes was observed in the θ-phase dominant sample and was unambiguously assigned to the θ-Al2O3 phase. The distinct line shapes related to the θ-Al2O3 phase provided an opportunity for effectively deconvoluting the more complex spectrum obtained from the δ-Al2O3 dominant sample, allowing the peaks/quadrupolar parameters related to the δ-Al2O3 phase to be extracted. The results show that the δ-Al2O3 phase contains three distinct AlO sites and three distinct AlT sites. This detailed Al site structural information offers a powerful way of analyzing the most complex γ-Al2O3 spectrum. It is found that the γ-Al2O3 phase consists of Al sites with local structures similar to those found in the δ-Al2O3 and θ-Al2O3 phases albeit with less ordering. Spin-lattice relaxation time measurement further confirms the disordering of the lattice. Collectively, this study uniquely assigns 27Al features in transition aluminas, offering a simplified method to quantify complex mixtures of aluminum sites in transition alumina samples.

Introduction

Transition aluminum oxides comprise an important class of materials broadly used across industries as catalysts, catalyst supports, adsorbents, hard protective coatings, abrasives, and membranes.[1−7] They impart such wide utilization due to their attractive structural, surface, and dielectric properties.[8] The structural stability, including the ability to maintain high surface area at elevated temperatures, is a key consideration solidifying the relevance of these materials in catalytic applications. Among the phases, γ-Al2O3 is the most prominent polymorph in catalytic applications due to its high surface area and structural stability up to the temperatures of 700–800 °C. However, after prolonged exposure to temperatures exceeding 800 °C, γ-Al2O3 transforms into δ-Al2O3 and θ-Al2O3 (∼1000 °C) polymorphs. At even further extended times and temperatures in excess of 1200 °C, the alumina finally converts to α-Al2O3, the most thermodynamically stable Al2O3 structure.[9,10] A key consequence associated with these phase transformations is a dramatic reduction in surface area, which deteriorates catalytic properties. For this reason, considerable experimental and theoretical investigations have been devoted in the past to understand the structural properties of γ-Al2O3 and the thermodynamic and kinetic factors that dictate its stability. It is now becoming well-understood that stabilization of transition Al2O3 with respect to α-Al2O3 can be achieved for very small crystallites as a consequence of low surface energy.[11] Extensive work has also focused on how to enhance the structural stability of the particle and understand the underlying mechanism of sintering. For example, it is found that by doping BaO or La2O3 onto the surface, the thermal stability of γ-Al2O3 is dramatically improved.[12,13]

Although significant advances have been made in the field of transition Al2O3,[14,15] the nature of the phase transformation at the early stages, γ-Al2O3 transforming to δ-Al2O3 and subsequently to θ-Al2O3, is the least understood. It should be noted that phase transitions that bypass δ-Al2O3 have also been observed.[16] A key reason for this lack of understanding is the ambiguity in the structure of γ-Al2O3,[17−23] which accommodates substantial defects to the microstructure, as well as the complex crystallographic structure of δ-Al2O3, which can also accommodate significant disorder.[24−27] Such δ-Al2O3 has been described as a complex intergrowth of two variants δ1 and δ2;[25,27] however, the nature of such a δ-Al2O3 phase has been shown to be more complex. δ-Al2O3 is actually described as adopting two intergrowth structures that are differentiated by their variant selection and intergrowth direction, δ1,2 and δ2,3,4.[26] To compound this issue, the coexistence of different transition Al2O3 phases in the same sample during heating has made it very challenging to probe the fine and irregular structures by traditional analytical techniques. One novel method to address this issue is to employ a recursive stacking methodology, accounting for the intergrowth modes of δ-Al2O3 and θ-Al2O3.[28] This is complicated by the amorphous domains, which are generated at the interfaces of ordered regions in the alumina crystals and impart disorder that cannot yet be fully accounted for with such a technique. Traditional XRD techniques, for example, require long-range order to understand the structure of alumina, which is complicated by the presence of a significant quantity of disordered regions. If crystals are free of strain and faulting, then peak broadening is only ascribed to domain size, which can be calculated by the Scherrer equation.[29] However, if the structure is amorphous or the nanoparticle size is smaller than 2 nm, then low signal to noise ratios and decreased detector sensitivity make the detection difficult and uncertain. The state-of-art XRD limitation is 1.2 nm on Au particles supported on carbon by using high sensitivity silicon slit detectors.[30] Chauhan et al. published a review on the XRD technique and its applications, demonstrating that a periodic array with long-range order is needed for diffraction to occur while the amorphous materials without long-range order do not display any significant peak in the diffraction pattern.[31]

Solid-state 27Al nuclear magnetic resonance (NMR) is a powerful technique for investigating the structural transformations and surface structure of alumina.[32−37] In particular, distinct NMR peaks are obtained for the tetra-, penta-, and octahedral aluminum sites present in various transition Al2O3 phases, as Chandran et al. recently reviewed.[38] Such an identification of the type of alumina present extends to the amorphous regions where other techniques, in particular for XRD, would fail due to the lack of long-range order since NMR is impacted primarily by the local environment. 27Al is a quadrupolar nucleus with a spin quantum I = 5/2 and is thus subject to quadrupolar line broadening. Because such line broadening is inversely proportional to the external magnetic field strength,[39] the resolution in magic-angle spinning (MAS) 27Al spectra of γ-Al2O3 obtained at low to medium magnetic fields is relatively poor, hindering clear separation of the peak corresponding to the penta-coordinated aluminum ions from those corresponding to the tetrahedral and octahedral aluminum. As such, the use of MAS NMR to study the relationship of the transition Al2O3 surface structure to its material and catalytic properties has been correspondingly limited. Only recently was high magnetic field NMR used to study the phase transition process of γ-Al2O3, with much improved spectral resolution over low magnetic fields for characterizing different alumina phases during the calcination.[40] A higher magnetic field is of great importance to provide better resolution for Al-containing compounds, such as higher resolution in an HBEA zeolite study[41] and identification of penta-coordinated Al sites in γ-Al2O3.[42] Our prior studies, conducted at a high field of 21.1 T, were the first to highlight the high-quality spectra of γ-Al2O3 with clearly separated tetra-, penta-, and octahedral aluminum sites.[13,42]

Multiple-quantum (MQ) MAS 2D NMR techniques[43,44] offer enhanced spectral resolution along the second dimension of a 2D NMR spectrum. The application of this technique to clearly resolve and assign spectral features to aluminum sites in tetra-, penta-, and octahedral coordinations was previously demonstrated.[45−47] Despite its remarkably rapid acceptance as an important technique for structural studies of a broad range of solid materials, especially microporous solids, glasses, and novel metal oxides, the MQ MAS experiment has aspects that still require further development. MQ MAS is, in general, not quantitative as there is no established method to allow uniform excitation and detection of multiple-quantum coherence that is independent of the magnitude of the quadrupolar coupling. In contrast, the very sensitive 1D MAS spectra (obtained at high fields and high spinning rates) are intrinsically quantitative, given appropriate experimental conditions. In these cases, the second-order line broadening is not eliminated but rather progressively narrowed at increasing magnetic fields.

In this work, three specially synthesized transition Al2O3 samples where the γ-, δ-, or θ-Al2O3 dominates are studied using a combined 1D 27Al MAS NMR and 2D MQ MAS experimental approach and coupled with spin–lattice relaxation techniques at a high field (19.9 T). This work assigns the observed spectral features to Al sites, which are present in the θ-Al2O3 and δ-Al2O3 phases, including the γ-Al2O3. The NMR parameters of different aluminum sites are extracted from the 2D MQ MAS NMR spectra and used to fit the signals obtained from the quantitative 1D MAS spectra, giving rise to quantitative distributions and spin–lattice relaxation times for different transition Al2O3 sites. To our knowledge, this is the first attempt to understand the aluminate sites in γ-Al2O3, where the advantages of the distinct line shapes from specially synthesized δ- and θ-Al2O3 dominated phases are utilized to gain deep insights into the aluminate site structures of the very complex γ-Al2O3 sample. Note that NMR is a powerful local structural probe where the local site structures can be probed, in particular the amorphous phase that can be especially emphasized by using a short recycle delay time due to its much shorter spin–lattice relaxation time compared to the crystalline phases, offering distinct advantages over the commonly used XRD technique.

Results and Discussion

The previously reported XRD patterns of the prepared alumina samples show that multiple phases of alumina exist in each of the samples.[25] The XRD pattern of the 800 °C calcined sample, γ-Al2O3 (Figure S1a), primarily consists of diffraction peaks consistent with γ-Al2O3 but are accompanied by a peak at 2θ = 46.4° indicative of δ-Al2O3, albeit visible with low peak intensity. The 1000 °C calcined sample, δ-Al2O3 (Figure S1b), exhibits features consistent with δ-, θ-, and γ-Al2O3 phases, where the sharp peaks at 2θ values of 46.4 and 32.6° are characteristic peaks for δ-Al2O3 crystals. Simulations of the overall XRD pattern without considering recursive stacking revealed that about 72% of the samples in crystallite forms in this sample are from the δ-Al2O3 phase (Table ).[25,27]Figure S1c reflects a higher abundance of θ-Al2O3 (69%) as is obvious by the group of sharp peaks at 2θ values of about 31.8, 32.6, 44.8, 47.5, 51, 60, 62.5, 64, 66, and 67°. These three samples are labeled γ-, δ-, and θ-Al2O3, respectively, to emphasize the dominant phase present in each sample. These samples correspond to crystallite sizes of 7 ± 2, 15 ± 1, and 23 ± 4 nm, respectively. The quantitative estimations of the fractions of alumina polymorphs in the nanoparticles using XRD are often difficult to generalize for metastable alumina phases that are not well-distinguished from the XRD patterns and characterized by complex intergrowths of variants. Furthermore, standard XRD pattern acquisition requires periodicity of the lattice to obtain a diffractogram, a requirement not met under highly amorphous conditions. For XRD to generate a peak, a particle with periodic lattice structure and dimensions greater than 2 nm is required. Complementarily, NMR is sensitive to local structural changes and does not require such a long-range order. Herein, we attempt to deconvolute and quantify the types of sites present in each prepared sample. It should be emphasized that different alumina phases are concepts of XRD analysis only, while 27Al NMR gives the aluminum site distributions in the octahedral, tetrahedral, and penta-Al site regions. Rigorous and quantitative correlation between NMR and XRD is impossible given the fact that surface aluminum sites such as penta-coordinated aluminum sites are not detected. Though XRD gives a signal that is a function of all atoms in the sample (including disordered and amorphous structures), the disordered and amorphous regions are usually disregarded, and XRD analysis typically looks at the ordered fraction of the domain with dimensions of greater than 2 nm. Thus, the current NMR analysis complements XRD studies and offers new insights into the aluminum site distribution.

Table 1

XRD Results for the Prepared Al2O3 Samples

sample	XRD determined phase ratio	crystallite size (nm)
γ-Al2O3	100% γ-Al2O3	7 ± 2
δ-Al2O3	52% δ1-Al2O3	15 ± 1
	20% δ2-Al2O3
	28% θ-Al2O3
θ-Al2O3	31% δ1-Al2O3	23 ± 4
	69% θ-Al2O3
	0.6% α-Al2O3

Figure presents the high-resolution solid-state 27Al MAS NMR spectra of these alumina samples. Figure a, obtained from the γ-Al2O3 sample, is similar to previously reported spectra.[13,42] The spectrum consists of three major peaks centered at 12.8, 35, and 70.6 ppm due to Al3+ cations in octa- (AlO), penta- (AlP), and tetrahedral (AlT) coordination, respectively. It should be emphasized again that γ-Al2O3 is purely an XRD identification based on the fairly broad XRD diffraction peaks in Figure S1a (i.e., lacking fine diffraction peaks). It is generally accepted that the structures in γ-Al2O3 are very complex and are also poorly understood in the literature. In this work, we will be able to provide new insights into the aluminum site distributions in the γ-Al2O3 sample by taking advantage of the specially synthesized δ-Al2O3 and θ-Al2O3 phase dominated samples. In the δ-Al2O3 sample (Figure b), two distinct AlO features emerge, centered at ∼10.2 and ∼12.8 ppm. Careful examination of the spectral features in Figure a reveals that these two AlO features are also present in the γ-Al2O3 sample albeit the peak linewidth is broader. The AlT shoulder peak feature at 64.5 ppm, a break point of a quadrupolar line shape as will be made clear later, is clearly visible in both Figure a and Figure b. In the θ-Al2O3 spectrum (Figure c), the octahedral AlO peaks are dominated by a strong resonance located at about 9.7 ppm. Further, the intensity of the shoulder at 64.5 ppm is weaker than that of δ-Al2O3 but still visible. An additional peak with a well-defined quadrupolar second-order line shape with a dominant break point located at about 74.3 ppm stands out in θ-Al2O3. According to the XRD results, this sample exhibits a dominating θ-Al2O3 phase (69%). As such, the feature peaks present at 9.7 and 74.3 ppm are safely assigned to the octa- and tetrahedral aluminum sites, which are associated with the θ-Al2O3 phase. This is consistent with previous assignments by O’Dell et al., where the 27Al MAS NMR spectrum was acquired at a lower magnetic field of 14.1 T.[34] Given the assignments of the AlO and AlT peaks for the θ-Al2O3 phase in Figure c, the sets of the peak features at 10.2 ppm (AlO) and ∼74 ppm (AlT) in the δ-Al2O3 dominated sample (Figure b) can be assigned to the AlO and AlT sites associated with the θ-Al2O3 structure. Since Figure b contains 72% of the δ-phase based on XRD, the major spectral features must be associated with the δ-phase. Thus, the dominant AlO peak centered at ∼12.8 ppm and the dominant AlT peak feature located at about 70.6 ppm (a break point of quadrupolar line shape) in Figure b are attributed to the spectral features associated with δ-Al2O3. With a close examination of the peak line shapes in both the octahedral and the tetrahedral spectral ranges in Figure , it becomes apparent that γ-Al2O3 consists of all spectral features associated with the δ-Al2O3 and θ-Al2O3 samples albeit the corresponding spectral features are broadened and the distribution of the Al sites is different. The increased line broadening is a direct consequence of the considerably increased disordering of the polycrystal structures, reflected by the significantly reduced crystallite sizes (7.5 nm for γ-, 15 nm for δ-, and 23 nm for δ-Al2O3) from XRD measurement.

Figure 1

1D 27Al MAS NMR spectra with a recovery time of 80 s for the three transition Al2O3 samples (a) γ-Al2O3, (b) δ-Al2O3, and (c) θ-Al2O3 (inserted are ×10).

The fact that γ-Al2O3 contains aluminum sites (AlO and AlT) similar to δ-Al2O3 and θ-Al2O3 is quite reasonable since the peak position in an NMR spectrum is only sensitive to the local electronic structure surrounding the observed Al site (i.e., up to about 1 nm), and δ-Al2O3 was previously described by a superstructure of γ-Al2O3 with the c-axis tripled because of the increased ordering of cationic vacancies on octahedral sites.[10,48] A similar local environment around Al3+ will result in the same peak position associated with these two Al2O3 phases. It is also reasonable considering the diverse sites identified for δ-Al2O3, which is composed of four variants that likely broaden the effective spectral features for the phase and contribute to a signal composed of several species with similar environments. Based on TEM studies, δ1-Al2O3 alone presents 10 unique octahedral sites and 6 unique tetrahedral sites, highlighting the potential for further ambiguity in peak position. Among all variants, there exist 57 unique sites in delta, in addition to the disordered transition zones (δ1: 10, 6; δ2: 10, 6; δ3: 6, 3; and δ4: 10, 6 for octahedral and tetrahedral, respectively).

It becomes apparent that the 12.8 ppm peak in Figure b is narrower than the corresponding 12.8 ppm peak in Figure a primarily due to the increased ordering in the δ-Al2O3 phase compared with γ-Al2O3. It is known from the XRD studies summarized in Table that the crystallite sizes for the samples increase with thermal treatment, showing a higher degree of order at samples expressing higher contents of later phases. Further, the intensity of the AlP sites was previously proposed as a reflection of the amount of disorder in aluminum samples[49] since no AlP sites exist in the idealized crystalline Al2O3 phases. From γ- and δ- to θ-Al2O3 dominant samples, the intensity of AlP sites decreases, showing that the structure is getting more and more ordered, consistent with the crystal size evolution. Compared with γ-Al2O3 and δ-Al2O3, the intensity of the tail on the low AlO ppm side (reflected by the red dashed line) in θ-Al2O3 is significantly decreased. The more symmetric line shape in θ-Al2O3 indicates that the structure is more ordered in this sample. The 27Al experiments conducted at a high magnetic field thus offer insights into the material structure changes that occur across thermal treatments, indicated by changes to the resonance feature evolution.

The results of 27Al 3Q MAS NMR provide new insights into the nature of the electronic environment around Al3+ cations in these nanostructured transition aluminas. Examples of sliced spectra from 3Q MAS NMR on γ-Al2O3 are also reported previously.[3] As can be seen in Figure b,c, 2D 3Q MAS NMR spectra of these two specially prepared nanostructure transition aluminas show significantly enhanced spectral resolution. This allows at least three different aluminum sites to be unambiguously detected in both the octahedral and tetrahedral coordinated environments in these two samples. In the octahedral region, on the AlO(1) site, which was not well-resolved in the 1D MAS spectra, a feature is distinguished in the 2D experiment. The absence of AlP sites in all of the 2D experiments shown here is due to their low abundance and an insufficient quantity of the accumulation number in data acquisition. This is no surprise as it is well-known that 2D 3Q MAS experiments offer the highest possible spectral resolution but at the expense of sensitivity and quantitative information when compared with 1D MAS experiments. Despite the lower sensitivity and the loss of quantification compared with quantitative 1D NMR, the isotropic chemical shifts (δiso), quadruple coupling constants (CQ), and asymmetry parameters (η) can be obtained from 3Q MAS NMR spectra (Table ), offering unique information for characterizing the different Al sites. These parameters were obtained by fitting 1D spectra using the DMFIT program on sliced spectra that were extracted parallel to the F2 (acquisition) dimension and at various F1 (isotropic) dimensions. These same parameters were then used to deconvolute the 1D spectrum (Figure d–f) as a strategy for obtaining quantitative information of different aluminum sites (Table ). It is necessary to use an additional octahedral resonance peak, AlO(4), in all three samples in order to fit the spectrum.

Figure 2

2D 27Al 3Q MAS NMR spectra of (a) γ-Al2O3, (b) δ-Al2O3, and (c) θ-Al2O3 with a 5 s recycle delay; (d–f) corresponding 1D MAS NMR spectrum (top) of them and their simulated spectrum (bottom) using the Q mas model in DMFIT (the red line in the middle is the sum of the individual fitting lines). Pentahedral sites are not simulated due to their low intensity.

Table 2

Spectroscopic Parameters Including Isotropic Chemical Shifts (δiso), Quadruple Coupling Constants (CQ), and Asymmetry Parameters (η) of Al Peaks and Their Relative Integrated Intensities in the Three Samples

sample	peak (Al site)	δiso (±1 ppm)	CQ (±0.1 MHz)	η (±0.1)	intensity (%)	Al site structure assignments	alumina phase %
γ-Al2O3	AlO(1)	16.9	4.9	0.0	3.6	local structures similar to δ
	AlO(2)	15.7	5.5	0.4	42.8	local structures similar to δ
	AlO(3)	11.0	4.4	0.5	9.2	local structures similar to θ	100% γ-Al2O3
	AlO(4)	8.6	6.0	0.6	7.9	disordered, local structures similar to δ
	AlT(1)	78.8	6.4	0.5	8.4	local structures similar to θ
	AlT(2)	73.9	6.1	0.5	22.4	local structures similar to δ
	AlT(3)	67.5	6.6	0.6	5.7	disordered, local structures similar to δ
δ-Al2O3	AlO(1)	16.3	4.8	0.0	17.0	δ
	AlO(2)	14.5	4.3	0.6	22.3	δ
	AlO(3)	10.8	3.5	0.5	12.0	θ	77.1% δ
	AlO(4)	9.8	5.9	0.6	11.3	disordered δ	22.9% θ
	AlT(1)	80.0	6.0	0.7	10.9	θ
	AlT(2)	73.2	4.6	0.6	12.9	δ
	AlT(3)	68.3	6.6	0.4	13.6	disordered δ
θ-Al2O3	AlO(1)	15.8	4.0	0.0	11.2	δ
	AlO(2)	13.2	2.9	0.3	11.5	δ
	AlO(3)	10.2	3.5	0.5	33.8	θ	37.3% δ
	AlO(4)	9.7	5.9	0.6	1.4	disordered δ	62.7% θ
	AlT(1)	80.3	6.2	0.7	28.9	θ
	AlT(2)	73.0	4.9	0.5	7.3	δ
	AlT(3)	66.7	6.6	0.4	4.5	disordered δ

Based on the results from 1D and 2D NMR, it is apparent that all three samples are complex mixtures of different alumina phases, in agreement with the XRD results discussed earlier. However, rigorous assignment of each individual peak to an XRD-identified phase is difficult due to the fact that XRD is a method based on long-range order, while NMR is a local structure-sensitive structure probe with a detection limit not reliant on periodicity. Thus, only those 27Al peaks associated with long-range order can be safely associated with an XRD-identified phase, while the peaks associated with more disordered environments, in particular the amorphous phase, can only be assigned to specific new Al sites. Herein, we will only attempt to establish a rough correlation between NMR and XRD results. Based on 1D and 2D NMR results and previously reported NMR parameters of different aluminum species/sites, assignments can be made for the individual simulated peaks, which are summarized in Table . The AlO(1) site has a chemical shift (16.9 ppm) near the chemical environment of the α-Al2O3 phase.[50] However, no α-Al2O3 phase was detected by XRD in the γ- and δ-Al2O3 samples (Figure S1a,b), and the temperature of treatment was too low for this stable phase to form to a significant extent.[17,24] In addition, the XRD-reported 0.6% for the α-Al2O3 phase in the θ-Al2O3 dominant sample would be too low to be observed by current 27Al 1D and 2D NMR. This Al3+ cation exhibits an integrated intensity, which is modulated from 3.6 to 17% and then to 11.2%. Such a decrease would be unexpected for α-Al2O3 phase formation given the better crystallinity of the θ-Al2O3 sample overall and the higher propensity to generate α-Al2O3 under such conditions. Since the abundance is highest in the samples that preferentially express the δ-phase dominant sample, this signal is assigned to octahedral aluminum sites in Al2O3, which are associated with one of the many δ-phase structures that have a local electronic structure (within 1 nm) similar to that of the α-Al2O3 octahedral site given the similarity of the experimentally observed chemical shift value of 16.9 ppm to that of α-Al2O3.

Based on the presence of the AlO(2) and AlT(2) features across all samples with decreasing abundance (42.8/22.4%, 22.3/13.0%, and 11.5/7.3% for AlO(2) and AlT(2), respectively), these sites are assigned to aluminum sites associated with the δ-Al2O3 phase for the δ-Al2O3 and θ-Al2O3 dominated samples (Figure b) since there are no γ-Al2O3 phases detected in these two samples. Then, the question arises on what the AlO(2) and AlT(2) features in the γ-Al2O3 sample belong to. The ratio of AlO(2)/AlT(2) determined from the experimental data is 1.91, 1.72, and 1.57 for γ-Al2O3, δ-Al2O3, and θ-Al2O3 dominated samples, respectively. Based on XRD studies, it is commonly accepted that the oxygen lattice for γ- and δ-Al2O3 phases is the same; the difference is in the ordering of the Al atoms. It has been suggested that aluminum cation rearrangement occurs and that vacancies are distributed among octahedral (Oh) and tetrahedral (Td) sites in γ-Al2O3 but ordered solely on octahedral sites in δ-Al2O3.[48,51,52] In general, defective spinel-like models are used to describe their structures, though departures from the spinel representation of the structure with the occupation of non-spinel sites have also been proposed.[17,19−23] In a spinel structure, the Al3+ distribution can vary between 62.5% Oh and 37.5% Td (Oh/Td = 1.66) when assuming that vacancies occupy exclusively Oh interstices[19] and 75% Oh and 25% Td (Oh/Td = 3) when vacancies are supposed to be exclusively located in Td sites. Thus, the ratio of AlO(2)/AlT(2) in the three samples to some extent agrees with the proposal that vacancies are distributed randomly in the γ-Al2O3 phase and ordered almost exclusively on Oh sites in the δ-Al2O3 and θ-Al2O3 phase dominant samples. Note that the ratios of AlO(2)/AlT(2) in our δ- and θ-Al2O3 samples are close to that of ideal δ-Al2O3 (1.66). This explains why the amount of the γ-phase is not detected by XRD (Table ). We can, however, note some differences by comparing the ratio of AlO/AlT for the two phases as well as the CQ values of AlO(2) and AlT(2) sites. CQ in the δ- and θ-Al2O3 samples is smaller than that of γ-Al2O3 and low-field literature reports of 5.5 MHz,[36] indicating that the structures of AlO(2) and AlT(2) sites are more ordered (i.e., with a longer range of periodic structures) in the δ- and θ-Al2O3 samples, consistent with XRD results.

Two additional Al sites, AlO(4) and AlT(3), reflect relatively high CQ values that are apparently associated with more disordered structures due to less symmetry of the electric structures around the corresponding Al sites. Compared to the AlO(4) and AlT(3) in the γ-Al2O3 dominated sample (7.9 and 5.7%, respectively), the intensities of the AlO(4) and AlT(3) sites increased in the δ-Al2O3 dominated sample to 11.3 and 13.6% and then decreased dramatically in the θ-Al2O3 dominated sample to 1.4 and 4.5%. Since no γ-Al2O3 is detected by XRD in both the δ-Al2O3 and θ-Al2O3 dominated samples (Table ), the AlO(4) and AlT(3) sites are then assigned to be associated with the δ-Al2O3 phase as these sites are the most prominent in the δ-Al2O3 dominated sample. Taking these new assessments in hand where AlO(1), AlO(2), AlO(4), AlT(2), and AlT(3) are all related to Al sites in δ-Al2O3, the ratio of AlO/AlT in the three samples is 1.93, 1.90, and 2.03. These values are closer to an Oh/Td of 1.66 than of 3, maintaining the previous assessment of δ-Al2O3 with vacancies occupying predominantly Oh interstices.

The relative intensity of AlO(3) and AlT(1) sites across all samples is almost equal within experimental uncertainty, (1.10, 1.10, and 1.12 for the γ-Al2O3, δ-Al2O3, and θ-Al2O3 samples, respectively, see Table ). This is evidence that AlO(3) and AlT(1) are related to each other. The fact that the relative intensity of these features increases between the γ-Al2O3 and δ-Al2O3 samples and dramatically increases for the θ-Al2O3 sample up to 64% suggests that it is related to the θ-phase, especially considering that the XRD results suggest a 69% θ-Al2O3 phase. It is commonly accepted that the structure of the θ-phase is a structural isomorph of β-Ga2O3, with the aluminum cations equally distributed over octahedral and tetrahedral sites, giving rise to AlO/AlT = 1.0 in an ideal case.[10] This is the approximate ratio between AlO(3) and AlT(1). However, recent work has combined NMR and XRD methods to show that the actual fraction of tetrahedral sites in the complex microstructure of θ-Al2O3 is ∼42%.[27] Taking into account all sites that may reflect Al species in θ-Al2O3 (AlO(1+3) and AlT(1)), we can arrive at adjusted tetrahedral aluminum fractions proposed for θ-Al2O3 in these samples of 39.6, 45.8, and 40.2%, for γ-Al2O3, δ-Al2O3, and θ-Al2O3, respectively. These values match well the more robust description of the site composition of θ-Al2O3.

θ-Al2O3 is the final transition alumina phase before transformation to the final thermally stable α-Al2O3 phase. It is well-known that when transition alumina undergoes the conversion to α-Al2O3, significant grain growth occurs,[53] and θ-Al2O3 has been reported to transform to α-Al2O3 at a nanocrystal size between 10 and 20 nm.[54] Considering the size of our samples, 15 ± 1 and 23 ± 4 nm for the δ-Al2O3 and θ-Al2O3 dominated samples, respectively, it is reasonable that small domains with local octahedral Al site structure (<1 nm) similar to that of α-Al2O3, thus the similar 27Al chemical shifts, may have formed in the θ-Al2O3 sample, as suggested by XRD. Such minority species would contribute to the intensity of signals in the octahedral region and may well overlap with signals at AlO(1).

Based on the above discussions and the assignments of the various octahedral and tetrahedral Al sites in both the δ-Al2O3 and θ-Al2O3 dominant sample by 27Al MAS NMR spectra, it becomes obvious that the Al sites in the so-called 100% γ-Al2O3 sample defined by XRD, in fact, consist of Al site structures with the local electronic environment (including structures with sizes of ∼2 nm or less) similar to the structure units found in both the δ-Al2O3 and θ-Al2O3 phases (see Table for details). The notable differences are that the structures in γ-Al2O3 are less ordered, i.e., more disordered with the least long-range periodicity. Given the fact that there are numerous δ-Al2O3 phase structures identified by XRD, so are for the θ-Al2O3 phase structures, such complexity combined with the least ordering is most likely the major reason responsible for the relatively broad 27Al peaks in both the octahedral and the tetrahedral Al sites compared to those observed in the δ-Al2O3 and θ-Al2O3 phases. This may also be the reason why the XRD pattern is also the broadest among the three specially made samples (Figure S1a).

γ-Al2O3 is often regarded as a defective spinel structure; however, the true nature of the phase is not yet fully described. δ-Al2O3 is said to be composed of two intergrowths (δ1,2 and δ2,3,4), which are composed of the δ1, δ2, δ3, and δ4 variants. The first two variants contain a slight difference in structure, and the latter three have slightly different translations/rotations. In all, these variants are responsible for 57 unique Al sites.[26] In contrast, θ-Al2O3 is composed of an intergrowth of just two variants, β-Ga2O3 with motifs of a monoclinic phase related to δ3-Al2O3. As for Al local environments, the amorphous phase cannot be detected by the XRD technique. Thus, 1D and 2D NMR measurement was utilized to provide information of Al local environments. The results are detailed in Table . AlO(4) and AlT(3) are ascribed to the disordered δ phase. In γ-Al2O3, AlO(1), AlO(2), AlO(4), AlT(2), and AlT(3) are assigned to local structure similar to the δ phase, and AlO(3) and AlT(1) are related with the θ phase. The Al sites in γ-Al2O3 are all similar to the sites associated with the θ and δ phase albeit the long-range ordering is different, i.e., crystalline phases have long-range ordering while the amorphous phase has none.

Outlined below, spin–lattice relaxation times (T1) of the various AlO and AlT sites were measured using the standard saturation recovery method. For each spectrum acquired at a specific recovery time (Figure a–c), the isotropic chemical shifts and the CQ values of various AlO and AlT sites extracted from 2D experiments in Table are used to fit the 1D spectrum, allowing less than 5% of variation of the parameters during fit optimization. The relative integrated intensity of each site as a function of recovery times is obtained and simulated as illustrated in Figure d–f. The relaxation data for most of the Al sites cannot be fit using a single exponential function, so a double-exponential function with distinct relaxation times, i.e., a fast-relaxing component (T1fast) and a slow-relaxing component (T1slow), was employed. The measured relaxation times for the two components and their relative abundance are summarized in Table . Due to the restricted freedom of motion in solid-state NMR, increased molecular/segmental motion often means a shorter T1 value.[55] Thus, the fast-relaxing component is related to increased freedom of motion due to small cluster sizes or aluminum atoms in more disordered/less ordered structures. It can be deduced from the presented fractions that there are more disordered structures in the γ-Al2O3 sample and the fewest in the θ-Al2O3 samples, supporting the results obtained from the quantitative 1D NMR results discussed above. Furthermore, it is shown that the distorted structures with short relaxation time are selectively enhanced at short recovery times, validating that the fast-relaxing component is, indeed, a more disordering of the structures. For example, the pentahedral part of γ-Al2O3 is more apparent at short recovery times with its T1 = 0.02 s. The broad line peaks of δ-Al2O3 acquired at a recovery time of 0.15 s are visually quite similar to the features of γ-Al2O3 in both the octahedral and tetrahedral spectral regions. At longer recovery times, the intensity of the slow-relaxing component grows, with the octahedral component of δ-Al2O3 split into two peaks (AlO(2) and AlO(3)). Further, AlO(3) grows faster than AlO(2), which is also reflected by their measured relaxation times. As such, it is necessary to wait until the spins are fully relaxed in order to get a quantitative 1D spectrum. The results of the T1 measurements may indicate the following: (1) Comparison of the spectra of 10 ms, 30 ms, 50 ms, 100 ms, and 1 s of the γ-Al2O3 dominated sample (Figure S2) illustrates that for the γ-Al2O3 dominated sample, the line shape does not change much with the recovery time; (2) comparison of short recovery time spectra for δ-Al2O3 and θ-Al2O3 dominated spectra with that of the γ-Al2O3 dominated sample clearly shows similar spectral features in terms of both the overall peak line shape and the relaxation time. This again emphasizes that the short relaxing time spectra from both the δ-Al2O3 and θ-Al2O3 dominated samples contain Al site structures similar to those found in the γ-Al2O3 phase, meaning similar local electronic structures. The results further strengthen the concept discussed above that the γ-Al2O3 phase (an XRD distinction) is, indeed, a composition of Al sites with local structures similar to those found in δ-Al2O3 and θ-Al2O3 phases. Such observations are where NMR can make a significant contribution to understanding the structural properties of transition aluminas and where XRD suffers from challenges impacted by a lack of long-range order.

Figure 3

Spin–lattice relaxation NMR spectra. (a–c) 27Al MAS NMR spectra of the prepared alumina samples as a function of several selected recovery times. (a) γ-Al2O3 (inserted pentahedral aluminum of γ-Al2O3, the accumulation number used is 3000), (b) δ-Al2O3, and (c) θ-Al2O3. (d–f) Integrated 27Al MAS NMR spectral intensity of each site as a function of the saturation recovery time in (d) γ-Al2O3, (e) δ-Al2O3, and (f) θ-Al2O3.

Table 3

Spin–Lattice Relaxation Times (T1) of Each Site in the Three Samples, by Fitting the Intensity of Each Aluminum Site as a Function of Saturation Recovery Timea

	sample A (100% γ-Al2O3)		sample B (δ-Al2O3 dominant)		sample C (θ-Al2O3 dominant)
27Al NMR peak (aluminum sites)	T1slow/s (fraction)	T1fast/s (fraction)	T1slow/s (fraction)	T1fast/s (fraction)	T1slow/s (fraction)	T1fast/s (fraction)
AlO(1)	19.6(79%)	0.8(21%)	15.6(82%)	0.7(18%)	29.9(80%)	0.2(20%)
AlO(2)	7.6(77%)	0.1(23%)	9.7(74%)	0.1(26%)	11.4(83%)	0.1(17%)
AlO(3)	6.1(70%)	0.2(30%)	7.8(89%)	0.05(11%)	13.9(83%)	0.9(17%)
AlO(4)	12.7(77%)	0.3(23%)	10.8(80%)	0.2(20%)	1.2(100%)
AlT(1)	5.3(79%)	0.1(21%)	7.3(90%)	0.2(10%)	15.7(92%)	0.4(8%)
AlT(2)	11.9(72%)	0.4(28%)	16.6(82%)	0.2(18%)	14.5(79%)	0.2(21%)
AlT(3)	10.3(100%)		16.0(81%)	1.1(19%)	10.0(88%)	0.7(12%)

It is necessary for most of the sites to use a double exponential rise function to fit well with the experiment results. The corresponding T1 is divided into two parts: T1slow and T1fast, and their fractions are determined.

Conclusions

High-resolution 1D and 2D 27Al MAS NMR spectra of three alumina samples where the γ-, δ-, or θ-phase was dominantly expressed by controlling the synthesis were obtained at a high field of 850 MHz (19.7 T). Six aluminum sites were resolved in the 2D MQ MAS NMR spectra, and seven aluminum sites are required to fit the quantitative 1D MAS NMR spectra. The spectral assignments as well as the isotropic chemical shift values and the quadrupolar coupling constants related to these sites were subsequently determined. A set of dominant peaks (i.e., AlO(3) and AlT(1)) with well-defined quadrupolar line shapes was clearly defined in the θ-phase dominant sample and is unambiguously assigned to the θ-Al2O3 phase. The combination of the distinct line shapes and peak positions related to the θ-Al2O3 phase provided the opportunity for effectively deconvoluting the more complex spectrum obtained from the δ-Al2O3 dominant sample, allowing the peaks/quadrupolar parameters related to the δ-Al2O3 phase to be extracted with the aid of DMFIT. The results show that the δ-Al2O3 phase contains three distinct AlO sites (i.e., AlO(1), AlO(2), and AlO(4)) and two distinct AlT sites (i.e., AlT(2) and AlT(3)). This Al site structural information offers a powerful way of analyzing the most complex γ-Al2O3 spectrum (100% γ-Al2O3 distinction measured by XRD). It is found that the γ-Al2O3 phase consists of Al sites with local structures similar to those found in both the δ-Al2O3 phase and the θ-Al2O3 phase albeit with less ordering (i.e., more distorted). Spin–lattice relaxation time measurement further confirms the disordering of the lattice. For the θ-Al2O3 dominated sample, an additional AlT(4) site (CQ = 7.3 MHz, η = 0.6) with a fraction of 1.34% is found and is most likely associated with disordered regions near the θ-Al2O3 phase structure that has not yet been reported before due to difficulties in observation at lower magnetic fields. Collectively, this study uniquely assigns 27Al features (both octahedral and tetrahedral in nature), which serves to represent the types of Al species present in the θ-Al2O3 and δ-Al2O3 phases present in transition aluminas, offering a simplified method to quantify complex mixtures of alumina phases and sites in transition alumina samples.

Experimental Section

γ-, δ-, and θ-dominant Al2O3 samples, herein termed γ-Al2O3, δ-Al2O3, and θ-Al2O3, respectively, were synthesized from aluminum isopropoxide by the hydrolysis method as previously reported.[56] Approximately 10 g of aluminum isopropoxide was added to ∼50 mL of water with vigorous stirring at 80 °C for 1 h. The mixture was subsequently transferred to a 125 mL Teflon liner inside a Parr vessel, which was heated in an oven at 200 °C for 24 h. After cooling to room temperature, the powder was collected by filtration, washed extensively with distilled water, and dried at 100 °C to generate the parent boehmite. The powder of boehmite was then converted to γ-Al2O3 by calcination at 800 °C for 2 h, δ-Al2O3 by calcination at 1000 °C for 6 h, and θ-Al2O3 by calcination at 1100 °C for 3 h.

XRD measurements were carried out on a Philips PW3040/00 X’Pert powder X-ray diffractometer using Cu Kα radiation (λ = 1.5406 Å) in step mode between 2θ values of 15 and 75° with a step size of 0.02°/s. Data analysis was accomplished using Rietveld refinement with crystal structures reported earlier[25−27] using TOPAS v5 (Bruker AXS). Crystallite sizes were estimated from line broadening with instrumental breadths calculated using the fundamental parameter approach.

All 27Al MAS NMR experiments were performed at room temperature on a Varian-Inova 850 MHz NMR spectrometer, operating at a magnetic field of 19.9 T. The corresponding 27Al Larmor frequency was 221.412 MHz. All 1D spectra were acquired at a sample spinning rate of 20 kHz ± 1 Hz, using a commercial 3.2 mm pencil-type MAS probe. A saturation recovery pulse sequence with the saturation segment consisting of a train of hard 90° pulses (50 pulses) with the pulse width of 1.2 μs for each pulse and a pulse spacing of 5 μs was used to obtain spin–lattice relaxation times (T1), where the recovery time varies between 500 μs and 100 s. Then, during the recovery time, the signal gradually grows back as a function of the recovery time. The number of pulses and the pulse spacing were adjusted to meet the saturation requirement. Each spectrum was acquired using a total of 192 scans with a π/2 read pulse and an acquisition time of 25 ms. The 27Al 3Q MAS NMR spectra at selected recovery times between 0.005 and 30 s were obtained using a z-filter 3Q MAS pulse sequence by inserting the same saturation segment as the 1D at the beginning of the 3Q MAS sequence at a sample spinning rate of 20 kHz ± 1 Hz. The optimized pulse widths were p1 = 2.4 μs, p2 = 0.9 μs, and p3 = 10.0 μs. In the hypercomplex 3Q MAS experiment, 96 transients for either the real and imaginary F1 FID were collected for each of the 90–192 evolution increments for the three samples. The acquisition time was 10 ms. Spectral widths for the F2 (acquisition) and F1 (evolution) dimensions were 100 and 40 kHz, respectively. All spectra were externally referenced to a 1 M Al(NO3)3 aqueous solution at 0 ppm. DMFIT was the software package used to simulate the spectra and fit the peaks.[57]