29Si{27Al}, 27Al{29Si} and 27Al{1H} double-resonance NMR spectroscopy study of cementitious sodium aluminosilicate gels (geopolymers) and gel–zeolite composites

The influence of starting materials and synthesis route on the properties and the structure of cementitious sodium aluminosilicate gels is not fully understood, partly due their amorphous nature and the fact that they often contain residual reactants, which can make the results of single-pulse NMR spectroscopy applied to these materials difficult to interpret or ambiguous. To overcome some of these limitations, 29Si{27Al} TRAPDOR NMR as well as 27Al{29Si} and 27Al{1H} REDOR NMR spectroscopy were applied to materials synthesized by the one-part alkali-activation route from three different amorphous silica starting materials, including rice husk ash. The latter led to formation of a fully amorphous sodium aluminosilicate gel (geopolymer), while the materials produced from the other silicas contained amorphous phase and crystalline zeolites. Application of the double-resonance NMR methods allowed to identify hydrous alumina gel domains in the rice husk ash-based material as well as significantly differing amounts of residual silica in the three cured materials. Four-coordinated Al existed not only in the aluminosilicate gel framework but also in a water-rich chemical environment with only a small amount of Si in proximity, likely in the alumina gel or possibly present as extra-framework Al in the aluminosilicate gel. The results demonstrate how the employment of different silica starting materials determines the phase assemblage of one-part alkali-activated materials, which in turn influences their engineering properties such as the resistance against chemically/biologically aggressive media.


Introduction
Cementitious sodium aluminosilicate gels (sometimes referred to as aluminosilicate inorganic polymers or geopolymers) can be produced by activation of sufficiently reactive low-calcium (alumino)silicate precursors, such as metakaolin, y ash or volcanic rock, with alkaline solutions, such as alkali hydroxide or alkali silicate solutions. 1 These materials possess benecial engineering properties in a range of 'cement-like' applications, i.e. as binders for mortars and concretes, particularly for applications where high resistance against chemical attack 2-5 or against high temperatures 6-8 is required. In addition, they possess characteristics that make them promising candidates for use as re-proong and refractory materials. [9][10][11][12] Cementitious sodium aluminosilicate gels are also extensively studied with regard to radioactive waste stabilization 13 and other applications that make use of their sorption properties. 14,15 However, despite intensive research regarding these various applications, their nanostructure and how it inuences their properties (e.g., their durability in cement-like applications 16,17 ) is not fully understood.
The current knowledge of the nanostructure of these sodium aluminosilicate gels and the inuence of their chemical composition on the former is based mainly on 29 Si single-pulse MAS NMR studies, complemented by 27 Al and 23 Na single-pulse MAS NMR data. [18][19][20][21][22][23] Additional information, in particular on the state of water in these gels, has been obtained by means of 1 H MAS NMR, 1 H- 29 Si CPMAS NMR as well as two-dimensional and multiple-resonance NMR methods. [23][24][25][26] The structure that emerges from these studies is an amorphous framework silicate with partial substitution of Al 3+ for Si 4+ in tetrahedral sites. Si in these gels is present in Q 4 (mAl) sites with m ¼ 1, 2, 3, and 4, the relative amounts of which depend on its overall SiO 2 /Al 2 O 3 ratio. Surface -OH groups occur only to a very limited extent in these gels. The negative framework charge caused by the Al 3+ -Si 4+ substitution is thought to be balanced by Na + ions and possibly by extra-framework Al (EFAL); the results of different groups differ regarding the locations and the state of hydration of the charge balancing Na + ions. Because of the above characteristics, cementitious sodium aluminosilicate gels are regarded as being closely related to the amorphous precursors of zeolite synthesis (and thus also related to crystalline zeolites to some extent), 27 albeit produced under water-decient conditions.
Very recently, a rened structural model has been put forward based on 27 Al, 23 Na, and 17 O 3QMAS NMR, 28 in which the charge balancing species are Na + ions coordinated to three framework oxygen atoms and three H 2 O molecules, Na + ions coordinated to four framework oxygen atoms and two H 2 O molecules, and six-coordinated EFAL. Differing from this proposal, an earlier study 25 that employed 27 Al 3QMAS and 27 Al { 1 H} REDOR 3QMAS NMR concluded that charge-balancing EFAL in sodium aluminosilicate gels is in four-fold coordination. Only in one case, 29 an attempt was made to examine the structural model in terms of Si-O-Al connectivity by means of a 29 Si{ 27 Al} double-resonance NMR method. In that study, 29 Si { 27 Al} REAPDOR NMR was employed to conrm the assignment of the resonances in the 29 Si single-pulse MAS NMR spectrum of a metakaolin-based sodium aluminosilicate gel to the different Q 4 (mAl) units, though these units were already well separated in the single-pulse spectrum.
In the present study, 29 Si{ 27 Al} TRAPDOR NMR, 27 Al{ 29 Si} REDOR NMR and 27 Al{ 1 H} REDOR NMR were used to facilitate accurate structural description of cementitious sodium aluminosilicate gels, produced by alkali-activation of rice husk ash via the so called 'one-part' synthesis route. 30 For comparative purposes, sodium aluminosilicate gel-zeolite composites produced by alkali-activation of other silica materials were studied too. It is demonstrated that the use of double-resonance NMR methods provides new insights into the phase assemblage and structure of these materials that is otherwise difficult to obtain. The obtained results highlight that the knowledge about conventional alkali-activated materials cannot be simply transferred to one-part alkali-activated cements without modi-cations, and in addition point to a convenient way to tune the properties of the latter materials.

Materials
The starting materials for the synthesis of the sodium aluminosilicate gel and the sodium aluminosilicate gel-zeolite composites (in the following referred to as 'composites') were sodium aluminate, a rice husk ash (denoted 'RHA'), a commercial microsilica (denoted 'MS') and a silica-rich by-product from chlorosilane production (denoted 'CR'). The chemical compositions of the starting materials are shown in Table 1. Except the sodium aluminate, all starting materials were almost completely amorphous, as determined by powder X-ray diffraction (XRD), and they had comparable specic surface areas in the range 20-50 m 2 g À1 . The sodium aluminate consisted almost completely of NaAlO 2 and contained minor amounts of NaAlO 2 $1.25H 2 O and natrite (Na 2 CO 3 ) as impurities; it had a median particle size of 17.3 mm, as determined by laser granulometry. A more detailed description of the starting materials has been provided elsewhere. 31,32 Pastes were obtained by rst dry-mixing the sodium aluminate with RHA, MS or CR in relative amounts to obtain a molar SiO 2 /Al 2 O 3 ratio of the mixture of $3.5 (i.e. Si/Al z 1.75) and subsequent mixing with water at a nominal water/solids ratio of 0.50. The resulting molar ratios of the pastes are shown in Table  2; sample designations refer to the silica starting material, the approximate SiO 2 /Al 2 O 3 ratio and the curing time of 1 day, as described in the following. The pastes were mixed in a planetary centrifugal mixer either for 3 min at 1750 rpm (MS_3.5_1d and CR_3.5_1d) or for 4 min at 1200 rpm (RHA_3.5_1d). Subsequently, the pastes were cast into cubic (20 mm Â 20 mm Â 20 mm) molds and cured in the open molds at 80 C and a relative humidity of $80% for 1 day in an oven with humidity conditioning. Aer this curing time, the hardened specimens were removed from the oven and the molds, allowed to cool down to room temperature and subsequently either tested for uncon-ned compressive strength or ground manually with mortar and pestle (agate) for further analyses. The compressive strengths were 17.7 MPa for MS_3.5_1d; 9.9 MPa for CR_3.5_1d; and 29.8 MPa for RHA_3.5_1d. The powdered samples for NMR analyses were stored in closed glass vials at laboratory temperature until required for testing.
As reported previously, 26,31 MS_3.5_1d and CR_3.5_1d yielded composites containing mainly zeolite Na-A [Na 4 (AlSiO 4 ) 4 -$9H 2 O] and an amorphous sodium aluminosilicate gel with a molar SiO 2 /Al 2 O 3 ratio of $2 (Si/Al z 1). In the case of MS_3.5_1d, some hydrosodalite [Na 6 (AlSiO 4 ) 6 $4H 2 O] was present in addition, while for CR_3.5_1d very minor amounts of a faujasite-type zeolite and/or zeolite Na-EMT were observed by XRD. The degree of reaction of the silica starting materials (MS and CR, respectively) in these composites was approx. 55-60%, meaning that the composites contained undissolved silica  particles too; the latter were found to have a partially leached, hydrated surface layer. For RHA_3.5_1d, it was found by XRD that the reaction products were completely amorphous, except minor amounts of thermonatrite (Na 2 CO 3 $H 2 O) that had formed due to slight carbonation of the alkaline paste. 32 In addition, the XRD patterns indicated that the degree of reaction of the silica starting material of RHA_3.5_1d was substantially higher than the degree of reaction of the starting materials of the composites (MS and CR). 32

NMR experiments
To obtain information about the chemical environment of Si and Al in the studied materials, 29 Si{ 27 Al} TRAPDOR NMR, 33,34 27 Al{ 29 Si} REDOR NMR and 27 Al{ 1 H} REDOR NMR 35,36 experiments were performed. These experiments have in common that rst a spin-echo experiment is performed on the nucleus under study (spectrum denoted S 0 ). Subsequently, the spin-echo experiment is repeated while a continuous wave radiofrequency (r.f.) pulse (TRAPDOR) or distinct 180 pulses (REDOR) are applied to a second nucleus whose inuence is to be investigated. This leads to attenuation of the intensity of the signal of the nucleus under study, depending on the spin-echo times and particularly for those sites which have the second nucleus in spatial proximity (spectrum denoted S). In the difference of the latter two spectra, DS ¼ S 0 À S, the sites with the second nucleus in proximity are emphasized. The ratio DS/S 0 is denoted TRAPDOR fraction or REDOR effect, respectively, and is a measure of the inuence (i.e. abundance and proximity) of the second nucleus on the nucleus under study. The magnitude of the TRADPOR fraction or REDOR effect for a specic site depends, however, not only on the abundance/proximity of the second nucleus but also on other structural parameters, such as the overall Si/Al ratio in the case of aluminosilicates. 37 29 Si{ 27 Al} TRAPDOR NMR experiments were performed on a Bruker DMX 400 spectrometer at 9.4 T, using a 7 mm tripleresonance probe and employing a sample spinning speed of 6.0 kHz. The 29 Si 90 and 180 pulse lengths were 9.8 ms and 19.6 ms, respectively. A 15 two-phase pulse modulation (TPPM) sequence 38 was applied for proton decoupling. The r.f. eld strength of the 27 Al TRAPDOR pulse was 25 kHz, corresponding to the selective 27 Al 90 pulse length of 10 ms for YAG's AlO 4 peak. Approximately weekly, measurements of the TRAPDOR pulse power were taken directly at the power amplier with a 60 dB attenuator and an oscilloscope; the measured pulse power was always in the range 125 AE 3 W. An echo time of 2s ¼ NT r ¼ 52T r ¼ 8.6667 ms (N: number of rotor revolutions; T r : rotor period) was applied. 4672 scans with a recycle delay of 90 s were accumulated for each specimen. Thermal effects that could differently affect the two parts of the experiment if no measures are taken, where excluded by alternating measurements with and without 27 Al TRAPDOR irradiation. 37 For RHA_3.5_1d, a 29 Si single-pulse MAS NMR spectrum was recorded on the same spectrometer. A sample spinning speed of 6.5 kHz and a 90 pulse of 7.5 ms duration were employed. The 15 TPPM sequence was applied for proton decoupling, and 128 scans with a recycle delay of 1200 s were accumulated. The recycle delay was chosen to ensure a quantitative measurement (recycle delay of at least ve times the longest relaxation time, T 1 , of all components) 39 despite the comparatively long T 1 that may occur for some species in the present samples. 31 27 Al{ 29 Si} REDOR NMR and 27 29 Si NMR spectra were referenced to kaolinite with its upeld resonance at À91.5 ppm, and all 27 Al NMR spectra were referenced to YAG with its AlO 6 resonance at 0.6 ppm. Deconvolution/tting of the spectra was performed with the SOLA module of the Bruker TopSpin soware, version 3.1. To facilitate unbiased tting of the double-resonance spectra, the S 0 spectrum and the S spectra for each echo time were summed for each sample, and the resulting sum spectrum was evaluated to obtain the chemical shi, the FWHM and the line shape (composed of Gaussian and Lorentzian) of the deconvoluted peaks. These parameters were then xed to obtain the intensities of the resonances separately for the S 0 spectrum and the S spectra for each echo time, respectively. To obtain the intensities of the resonances in the 29 Si single-pulse MAS NMR spectrum of RHA_3.5_1d, the chemical shis, FWHMs and line shapes determined for its 29 Si{ 27 Al} TRAPDOR NMR spectrum were adopted for the t of the single-pulse spectrum. The Q n (mAl) nomenclature for SiO 4 tetrahedra is used throughout this article, where n denotes the number of oxygen-bridges to neighboring SiO 4 and AlO 4 tetrahedra, and m # n denotes the number of AlO 4 of these tetrahedra; for m ¼ 0, the expression in parentheses is omitted.

Results and discussion
3.1 29 Si{ 27 Al} TRAPDOR NMR Fig. 1 shows the 29 Si{ 27 Al} TRAPDOR spectra of the three materials. The S 0 spectra of all materials were very similar to their respective 29 Si single-pulse MAS NMR spectra (not shown; for MS_3.5_1d and CR_3.5_1d, see ref. 26 and 31); thus, the Si speciation in these materials can be discussed on the basis of the former.
In the S 0 spectra of MS_3.5_1d and CR_3.5_1d, the broad resonances at approx. À110 ppm and À98 ppm are assigned to Q 4 sites in unreacted silica and Q 3 sites in its partially hydrated (leached) surface layer, respectively. The major peak with its maximum at approx. À89 ppm and its downeld shoulder, present in both composites, are attributed to Q 4 (4Al) sites in three different phases: The peak at À89 ppm is assigned to zeolite Na-A. Sample MS_3.5_1d contained hydrosodalite, which causes a resonance at À87 ppm, visible as a shoulder in its S 0 spectrum. The spectra of both composites exhibited a shoulder at approx. À85 ppm, which has been shown to represent an amorphous sodium aluminosilicate gel with a molar SiO 2 /Al 2 O 3 ratio of $2 (Si/Al z 1). 26,31 For CR_3.5_1d, the fraction of Si in that gel has been determined to be 11% of the total Si in the system. 26 The S 0 spectrum of RHA_3.5_1d exhibited an essentially feature-less, broad hump in the range À75 ppm to À120 ppm, i.e. in the full range of Q 4 and Q 4 (mAl) sites with m ¼ 1.4. This resembles typical cementitious sodium aluminosilicate gels, which generally contain a distribution of Q 4 (mAl) sites with m ¼ 1.4, generally with the resonances of the Q 4 (mAl) sites broadened so that their 29 Si single-pulse MAS NMR spectra display only a similar broad hump. [18][19][20][21][22][23] From deconvolution of the 29 Si{ 27 Al} TRAPDOR spectra, the TRAPDOR fractions (DS/S 0 ) of the Q 4 (mAl) sites for the three materials were obtained ( Fig. 2; Table 3). In the composites, the TRAPDOR fractions of the Q 3 and the Q 4 species were very low, viz. $5% for Q 3 and $0% for Q 4 . This is in agreement with the assignment of these sites to the leached surface layer and the inner regions of unreacted silica in the composites, respectively. The slightly raised DS/S 0 of the Q 3 sites can be explained by interaction of Si in the surface layer with Al in the surrounding aluminosilicates and with diffusion of a limited amount of Al into the surface layer, leading to interactions with Si residing further inside the hydrated layer. Thus, a well-dened interface with an abrupt change of composition is not present between the unreacted silica and the surrounding zeolites and gel; instead, it appears that there is a more gradual transition between the silica and the aluminosilicate matrix, suggesting that also some bonding exists between these.
The deconvolution of the 29 Si{ 27 Al} TRAPDOR spectra of RHA_3.5_1d yielded resonances at approx. À86, À91, À96, À102 and À109 ppm, assigned to Q 4 (mAl) with m ¼ 4, 3, 2, 1 and Q 4 , respectively (Tables 3 and 4). The Q 4 species are assigned to unreacted silica RHA. Quantication of the relative intensity of  that resonance in the 29 Si single-pulse MAS NMR spectrum of RHA_3.5_1d yielded an abundance of 11% (Table 4). That means that the silica RHA had reacted to a degree of 89% in the material, signicantly higher than the degree of reaction of the silicas in MS_3.5_1d and CR_3.5_1d (Section 2.1). As the silica starting materials had comparable specic surface areas, the faster reaction kinetics of RHA are assigned to a higher fraction of network-modifying elements (sum of Na, K, Mg and Ca) in the silica; it is also noted that RHA contained substantial amounts of phosphorus and sulfur, which may increase its reactivity too.
The Q 4 (mAl) sites with m ¼ 1.4 in RHA_3.5_1d represent the sodium aluminosilicate gel (geopolymeric gel), 20,21,29,40 i.e. the product of the reaction of the silica RHA with the sodium aluminate. From their relative intensities (I) in the 29 Si singlepulse MAS NMR spectrum, the molar SiO 2 /Al 2 O 3 ratio of the sodium aluminosilicate gel, excluding the unreacted silica RHA, can be computed with Engelhardt's formula, 41 provided that virtually no Al-O-Al bonds are present in the material (Loewenstein's rule). The latter assumption has been conrmed by 17  The TRAPDOR fractions of the Q 4 (mAl) sites in the sodium aluminosilicate gel in MS_3.5_1d, CR_3.5_1d, RHA_3.5_1d, zeolite Na-A and the hydrosodalite were in relative good agreement (Table 3), considering that their determination involved separate deconvolution for each of the materials. However, the slightly lower DS/S 0 of Q 4 (4Al) in the gel of RHA_3.5_1d, compared to the Q 4 (4Al) in the gels of the composites, may be partly caused by its higher overall SiO 2 /Al 2 O 3 ratio, as the TRAPDOR fraction is not only determined by next-nearest neighbor Al atoms (Si-O-Al bonds) but also by Al that resides further away from the nucleus under study (e.g., Si-O-Si-O-Al).
The TRAPDOR fractions of the Q 4 (mAl) sites in the gel of RHA_3.5_1d decreased with decreasing m in the expected order (Table 3). DS/S 0 of the Q 4 (1Al) sites in RHA_3.5_1d was determined to be 15%, i.e. considerably higher than that of the Q 3 sites in the composites but only slightly lower than the TRAP-DOR fraction of the Q 4 (2Al) sites in RHA_3.5_1d. This indicates that any Q 3 species in a leached surface layer of the unreacted silica RHA, if present at all, contributed only to a minor extent to the resonance assigned to Q 4 (1Al). This would also be in line with the high degree of reaction of RHA. Because of this, the above use of the intensity of the Q 4 (1Al) sites, without a reduction to account for possible Q 3 species, to calculate the SiO 2 / Al 2 O 3 ratio of the aluminosilicate gel is justied.

27 Al{ 29 Si} and 27 Al{ 1 H} REDOR NMR
RHA_3.5_1d contained signicant amounts of aluminium in four-fold coordination (AlO 4 ) as well as in six-fold coordination (AlO 6 ), as shown by its 27 Al REDOR NMR S 0 spectra (Fig. 3). The maximum of the six-coordinated Al was located at a chemical shi of 8 ppm. The four-coordinated Al caused a resonance with its maximum at 58 ppm and a shoulder, centered at $65 ppm. As the resonance of Al in sodium aluminate in 27 Al MAS NMR spectra is found at $77-78 ppm, 42,43 this proves that the sodium aluminate has dissolved and reacted virtually completely during curing of RHA_3.5_1d.
The signal of AlO 6 did not exhibit a discernable 27 Al{ 29 Si} REDOR NMR effect (Fig. 3a), while its 27 Al{ 1 H} REDOR NMR effect was determined to be 88% (Fig. 3b). This proves that the six-coordinated Al existed in a proton-rich (i.e. water-rich) environment with no or only little Si in proximity, strongly suggesting that it formed a hydrous aluminate phase. The occurrence of a separate aluminate phase in RHA_3.5_1d is in line with the above nding that the SiO 2 /Al 2 O 3 ratio of its sodium aluminate gel was higher than the overall SiO 2 /Al 2 O 3 ration of the starting mix and the complete reaction of the sodium aluminate. As no crystalline compounds were detected by XRD in RHA_3.5_1d (Section 2.1), this must be an amorphous aluminate, i.e. alumina gel. This interpretation is in accord with work of Brew and MacKenzie, 43 who detected minor amounts of AlO 6 in samples produced by addition of silica fume Table 3 Chemical shifts (d) and TRAPDOR fractions (DS/S 0 ) of the Si sites in MS_3.5_1d, CR_3.5_1d and RHA_3.5_1d, obtained from deconvolution of their 29 Si{ 27 Al} TRAPDOR NMR spectra

Material
Phase/site to sodium aluminate solution and assigned these to poorly ordered Al(OH) 3 . The same reasoning has been adopted previously to explain the occurrence of minor amounts of AlO 6 in MS_3.5_1d and related materials with other SiO 2 /Al 2 O 3 ratios. 31 The 27 Al{ 29 Si} and 27 Al{ 1 H} REDOR NMR results presented here conrm these previous assignments.
The two AlO 4 species at 58 ppm and $65 ppm, respectively, displayed different behavior regarding their coupling with Si and protons (Fig. 3). The major signal at 58 ppm exhibited a 27 Al { 1 H} REDOR NMR effect of 39%, i.e. lower than the effect of the AlO 6 sites, and a 27 Al{ 29 Si} REDOR NMR effect of 2.5%. The latter gure has to be considered as indicating abundant Si in spatial proximity, as only 29 Si isotopes (natural isotope abundance 4.7%) contribute to DS/S 0 . The shoulder at $65 ppm exhibited a 27 Al{ 1 H} REDOR NMR effect similar to that of the AlO 6 sites, indicating a water-rich environment, while a signicant 27 Al{ 29 Si} REDOR NMR effect was not observed.
A conrmation of the observations regarding the AlO 4 resonance at 58 ppm and the AlO 6 resonance at 8 ppm is obtained from a plot of the DS/S 0 versus echo time (Fig. 4). Already the faster decay with increasing echo times of the S 0 signal of the AlO 6 resonance, compared to the AlO 4 resonance, indicates shorter spin-spin relaxation times (T 2 ), which results from a higher water abundance around the former. This is conrmed by its much steeper increase of its 27 Al{ 1 H} REDOR NMR effect with echo time. Also, the maximum 27 Al{ 1 H} REDOR NMR effect of the AlO 6 resonance ($100%) was signicantly higher than that of the AlO 4 resonance (73% at 2s ¼ 16.00 ms). The 27 Al{ 29 Si} REDOR effect of the AlO 4 resonance increased approximately linear with echo time to 25% at 16.00 ms, while no Al-Si coupling could be detected for the AlO 6 resonance, even at long echo times.
From the above results, it is clear that the four-coordinated Al with its resonance at 58 ppm is Al in the framework of the sodium aluminosilicate gel, in line with previous 27 Al MAS NMR studies of these materials. 18,19,22,23,31,40 It has Si in close proximity (Al-O-Si bonds), and the abundance of water in its proximity is signicant (water of hydration of charge balancing ions and pore water in the gel) but less than for the Al in the alumina gel.
The AlO 4 sites with their resonance at $65 ppm, which have few or no close Si atoms (Al-O-Si bonds) but abundant water in proximity, could be thought to be four-coordinated extra-  framework Al (EFAL) species. The existence of four-coordinated EFAL in cementitious aluminosilicate gels has been proposed by Brus et al.,25 referring to the occurrence of such species in dealuminated zeolite H-Y. 44,45 The authors gave the 27 Al isotropic chemical shis (d iso ) of the framework Al and the EFAL in their aluminosilicate gel as $61 ppm and $69 ppm, respectively, reasonably close to the experimental chemical shis (d) of the two AlO 4 species in the present study. However, it may be noted that four-coordinated EFAL in dealuminated zeolite H-Y is not always stable in the presence of excess water: while it persists up to at least 80% relative humidity in steamed zeolite H-Y, 44 it transforms to six-coordinated EFAL at high water loadings in calcined zeolite H-Y. 46 An alternative explanation for the occurrence of two different resonances in the AlO 4 range of the 27 Al MAS NMR spectra of sodium aluminosilicate gels has been proposed recently. 28 In that study, resonances were found at isotropic chemical shis of $61 ppm and $66 ppm and assigned to framework Al balanced by Na + and framework Al balanced by six-coordinated EFAL, respectively. However, adopting this assignment for the two AlO 4 species in the present study would require that the species causing the signal at $65 ppm had abundant Si in proximity (i.e. framework Al), which is not the case here. Therefore, this explanation can be excluded for the present materials. Another possibility is that the signal at $65 ppm in RHA_3.5_1d is caused by a minor amount of AlO 4 sites in its hydrous alumina gel. For example, Isobe et al. 47 have identied AlO 4 with a 27 Al chemical shi of 65-68 ppm as well as AlO 5 species besides AlO 6 in amorphous aluminium hydroxide precipitated from AlCl 3 /NaOH solution, proving that not necessarily all Al in alumina gel is in six-fold coordination. From the present results, it cannot be conclusively decided whether this latter explanation (AlO 4 in alumina gel) or the possible explanation mentioned rst (fourcoordinated EFAL) applies to the sites with a resonance at $65 ppm in RHA_3.5_1d, but the above reasoning about the stability of four-coordinated EFAL in the presence of excess water may be taken as an indication in favor of AlO 4 as a minor constituent of the alumina gel.

Conclusions
29 Si{ 27 Al} TRAPDOR NMR, 27 Al{ 29 Si} and 27 Al{ 1 H} REDOR NMR experiments provide information about alkali-activated materials that is otherwise difficult to obtain and thus are useful methods to complement single-pulse MAS NMR studies. In particular, they allow to unequivocally identify amorphous byproducts such as amorphous alumina gel and they contribute to a reliable identication, and thus quantication, of unreacted amorphous precursors in these materials.
As for the materials studied here, 29 Si{ 27 Al} TRAPDOR NMR experiments have conrmed that the composites based on silica MS and silica CR contained substantial amounts of unreacted silica with a leached surface layer. 27 Al{ 29 Si} and 27 Al{ 1 H} REDOR NMR experiments revealed that the amorphous material based on silica RHA was not a phase-pure sodium aluminosilicate gel but contained coprecipitated hydrous alumina gel. The RHA-based material contained unreacted silica in addition, but its amount was markedly smaller than in the composites. The combination of the latter two methods also showed that two four-coordinated Al species occurred in the RHA-based material, one of which was the framework Al in the aluminosilicate gel, while the other was likely AlO 4 in the alumina gel or possibly four-coordinated extra-framework Al in the aluminosilicate gel.
The results demonstrate that the choice of the silica feedstock in the production of one-part alkali-activated materials has a very important effect on the phase assemblage of the cured products: though all employed silicas were amorphous, the materials synthesized from MS and CR were gel-zeolite composites containing a substantial amount of unreacted silica (i.e. 'excess' silica), while the materials synthesized from the rice husk ash RHA was a completely amorphous sodium aluminosilicate gel containing hydrous alumina gel as byproduct (i.e. 'excess' alumina). It is noted that occurrence of alumina gel is likely an advantage for the previously proposed application of one-part alkali-activated materials as binders for sewer repair mortars. 5 One of the most important aspects for this application is the resistance of the mortars against biogenic sulfuric acid attack. It has been proposed 48,49 that Al(OH) 3 in calcium aluminate cements increases the resistance against biogenic sulfuric acid attack of mortars and concretes produced from these cements by releasing bacteriostatic Al 3+ into solution and thus inhibiting the activity of sulfuric acid-producing bacteria on the surface of the mortar or concrete. This suggests that coprecipitated hydrous alumina gel in one-part alkali-activated materials, which was observed here for RHA_3.5_1d, would have the same effect.
The different phase assemblages of the materials based on the different silica starting materials can likely be explained by differences of the dissolution kinetics of the silicas under alkaline conditions, caused by differences of the amounts of network-modifying constituents. This suggests that choice of a silica starting material with suitable dissolution kinetics, adjusted to the overall SiO 2 /Al 2 O 3 ratio of the reaction mixture, will lead to virtually complete conversion of the starting materials into phase-pure sodium aluminosilicate gel, which can impart improved engineering properties, e.g., higher mechanical strength, to the material. It may also be possible to yield the same effect by mixing different silica starting materials, but this suggestion has to tested in future studies. If true, this would add another benet to one-part alkali-activated materials, which, due to the avoidance of the employment of highly alkaline activator solutions, have already substantial advantages in terms of safety and handling compared to alkali-activated materials produced by the conventional synthesis route. 30,[50][51][52]