See discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/228528408 Supermicroporous silica-based SiO2-Al2O3- NiO materials: Solid-state NMR, NMR relaxation and magnetic susceptibility ARTICLE in MICROPOROUS AND MESOPOROUS MATERIALS · SEPTEMBER 2009 Impact Factor: 3.45 · DOI: 10.1016/j.micromeso.2008.08.023 CITATIONS 8 READS 44 5 AUTHORS, INCLUDING: Vladimir I Bakhmutov Texas A&M University 351 PUBLICATIONS 2,771 CITATIONS SEE PROFILE Andrey Prosvirin Texas A&M University 97 PUBLICATIONS 2,794 CITATIONS SEE PROFILE Kim R Dunbar Texas A&M University 443 PUBLICATIONS 13,397 CITATIONS SEE PROFILE Abraham Clearfield Texas A&M University 471 PUBLICATIONS 13,691 CITATIONS SEE PROFILE Available from: Vladimir I Bakhmutov Retrieved on: 04 February 2016
Microporous materials SiO2–Al2O3–NiO with a pore size of 8–20 Å, prepared by the sol–gel method atwide variation in Ni2+ concentrations, have been studied by X-ray diffraction, X-ray photoelectron spec-troscopy, transmission electron microscopy, magnetic susceptibility measurements and multinuclearNMR. It has been shown that the MAS NMR spectra and particularly NMR relaxation in amorphous para-magnetic solids can be successfully used for their characterizations. The 1H and 29Si spin-lattice NMRrelaxation is always non-exponential and governed by direct dipolar interactions with paramagnetic cen-ters while the spin-diffusion mechanism is negligible. The 29Si relaxation rates, obtained in the limits ofthe stretched exponential, reflect distribution of paramagnetic centers. It has been found that aluminumatoms are incorporated into the silica matrix of the materials while nickel centers are accumulated withinpores. The nature of the nickel centers has been studied by magnetic susceptibility measurements sup-ported by the XPS and X-ray experiments. These centers represent NiO and Ni0 (observed at high nickelloadings) aggregated into nanoparticles. The Ni0 nanoparticles are responsible for the room-temperatureferromagnetic behavior of the materials prepared with high nickel loadings.
� 2008 Elsevier Inc. All rights reserved.
1. Introduction
Ever since the discovery of mesoporous materials, such as theMCM-41 family [1,2] intense research into the synthesis and char-acterization of related porous materials has followed [3]. By usingsurfactants with moderately long alkyl chain lengths, pores in therange of 20 to greater than 100 Å diameters have been prepared.While these materials are semi-crystalline, the pores are well or-dered. The well-known zeolites are more highly crystalline andcommercially valuable, but are limited by pore sizes or pore en-trances of about 10 Å diameter [4]. There is thus a gap in materialsthat are in the 10–20 or 25 Å range. We have utilized sol–gel tech-niques, with amines as templates, to prepare amorphous mixedoxides with pore sizes largely in between those of zeolites andthe mesoporous materials [5]. Previous reports on our mixed sil-ica–alumina–MO (M = Ni, Mn and Zn) families of materials withbroad range values of MO have been published [6–8]. Because ofthe amorphous nature of these materials it is difficult to determinewhether a single phase is present over this broad range of compo-sitions. Potentially such materials may be active as catalysts [6] orsorbents but because of their poor crystallinity structural informa-
ll rights reserved.
: +1 979 845 4719.. Clearfield).
tion via X-ray diffraction methods is not forthcoming. Thus, a keystructural question remains: are transition metal atoms uniformlyincorporated into the silica/alumina matrix of the mixed oxidematerials or are they situated within the cavities? In addition, or-ganic templates, applied in the sol/gel synthesis, can reduce metalions under high-temperature calcinations of the materials.
Solid-state NMR is one of the powerful physical methods widelyapplied in materials science [9,10]. This method is particularly suc-cessful when materials are diamagnetic. Even for the stronglyamorphous molecular systems, the various solid-state MAS NMRtechniques can show important structural details that are oftenunavailable by other physical methods. In contrast, applicationsof solid-state NMR for paramagnetic solids are often non-trivial be-cause of strong electron–nucleus coupling, intense dipole–dipoleelectron–nucleus interactions and also effects of the relaxationand BMS (bulk magnetic susceptibility) nature [7,11,12]. Big con-tact chemical shifts with large anisotropies [13–18] in combinationwith short relaxation times [19] lead to a situation where detectionof NMR signals is difficult and their interpretation is problematic.Generally the nuclei, which are located closely to paramagneticcenters, for example, in their first coordination sphere, are ‘‘invisi-ble” in the standard MAS NMR spectra while the nuclei, remotefrom these centers, show resonances accompanied by intense side-bands [11,17,20]. Additional difficulties are connected with the factthat short relaxation times and sideband effects can be observedeven in systems where paramagnetic ions have been mechanically
V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86 79
mixed with amorphous oxides [7]. Under these circumstances itseems to be very important to reveal what structural informationcan be extracted from the solid-state NMR data.
In our previously published papers we reported the synthesis oftwo series of supermicroporous mixed oxides ZnO–Al2O3–SiO2 [8]and MnO–Al2O3–SiO2 [7], by the sol/gel method, in which tetraeth-ylorthosilicate (TEOS) was the solvent, as well as the source of SiO2.Primary amines of different chain lengths were used as templates.In the case of the ZnO series the surface area increased and thepore size range became larger as the chain length of the amine in-creased for a fixed oxide composition. Although changes in the so-lid-state MAS NMR of the zinc-containing oxide mixture relative toa prepared baseline alumina–silica mixture indicated the presenceof Zn–O–Al and Si–O–Zn bonding, the homogeneity of the mixturecould not be determined. This was not the case for the manganesecontaining samples.
Paramagnetic ions have a pronounced effect on magnetic nucleidepending upon the proximity of the two species. By comparingthe SS MAS NMR spectra of a mixture of Mn(O2CCH3)2 and a sil-ica–alumina baseline composition with spectra of the MnO–Al2O3–SiO2 series in which MnO was increased from 1 wt% to20 wt%, information on the homogeneity was forthcoming [21].MnO was found to be homogeneously dispersed in the silica–alu-mina matrix up to about 2.5 wt%. At larger amounts the excessMn2+ migrated out of the mix to line the pore surfaces [7,21,22].
Previously we synthesized a series of NiO–SiO2–Al2O3 materialsand utilized them in reduced form as catalysts [6]. In this paper, wefocus on multinuclear NMR and on NMR relaxation and magneticsusceptibility measurements in order to establish the nature anddistributions of paramagnetic centers in systems where Ni2+ load-ings have been varied from 0 to 30 wt% in a silica–alumina matrix.
2. Experimental
Materials 1–8 (Table 1) were obtained by a sol–gel method byusing tetraethylorthosilicate (TEOS) and aluminum tri-sec-butox-ide (ATSB) as sources of silicon and aluminum, respectively.Nickel(II) acetate tetrahydrate was used as the source of nickeland cyclohexylamine was added as an organic template and a base.The approximately constant content of carboxyl ions was con-trolled by addition of benzoic acid. All reagents were of ACS gradefrom Aldrich. The synthesis procedure was very similar to that re-ported earlier in details discussed for the zinc-containing materials[8]. Typically, TEOS is placed into a beaker and completely homog-enized with ATSB. After the homogenization, the amine was addedfollowed by Ni(OOCCH3)2 � 4H2O and the mixture stirred for 45–50 min. Then, ddi H2O (distilled and deionized) was added to thebeaker as required. The amount of water to be added was calcu-lated by taking into account the four moles of water present permole of nickel acetate. After the whole was aged for 20 h, the bea-ker was placed into an oven and the temperature was gradually
Table 1The analytical data on nickel contents and NLDFT cumulative surface area and pore volumeas 2–3 wt%)
raised to 200 �C within 3–4 h. The gels were heated at this temper-ature for 24 h in air to dryness. The final products were calcined byovernight heating at 450 and 540 �C in air to remove the organiccomponents [6]. The samples are marked through the text as1(450)–8(450) and 1(540)–8(540), respectively.
The 29Si, 1H, 2H, 27Al, 13C MAS NMR experiments were per-formed with a Bruker Avance-400 spectrometer equipped withstandard 4 and 7 mm MAS, and wide-line probe heads. The 29Si,27Al Hahn-echo MAS NMR spectra were obtained with echo-delayscorresponding to rotor periods. Phase corrections in these spectrawere carried out at PHC0 = 0 and PHC1 = SWH � (echo delay) � 360.The 27Al MAS NMR spectra were recorded with low-power pulsesto excite selectively the central transitions. The 13C and 29Si MASspectra were also collected with a common cross-polarizationpulse sequence. The external standards used for the 27Al and 13C,29Si, 1H NMR experiments were [Al(OH2)6]3+ and TMS, respectively.A wide-line probe was used for the variable-temperature 2H NMRexperiments performed with a quadruple-echo pulse sequence.
The 29Si, 1H and 2H T1 relaxation measurements were preformedby the standard inversion–recovery (180�–s–90�) experimentswith accurate pulse adjustments because a bad adjustment, itself,can lead to an exponential relaxation process [19]. The collectedspectra were improved by manual phasing and baseline correc-tions. The relaxation inversion–recovery curves were treated usingthe appropriate nonlinear fitting routine. The T1 values are mea-sured with errors 610% and well reproduced by independentexperiments.
X-ray diffraction patterns were collected with a Bruker D8 dif-fractometer using CuKa radiation (1.5418 Å) at 40 kV and 40 mAand a diffraction beam graphite monochromator. The measure-ments were recorded in steps of 0.02–0.04� with a count time of1–2 s in the 2h range of 2–50–70�.
Surface area measurements were performed on an Autosorb-6(Quantachrome) unit with nitrogen absorption at liquid nitrogentemperature. Both, pure N2 as an absorbate and He as a carriergas were utilized. Pre-calcined samples were out-gassed at300 �C for 18 h. The resulting data were analyzed using the stan-dard Autosorb-6 software supplied by Quantachrome Corp. Surfaceareas were calculated on the basis of the BET model. The more de-tailed information on the cumulative pore volume and surface areadata, as well as pore size distribution characterization, were basedon the NLDFT calculations carried out by using the advance Auto-sorb-1 Version 1.51 package supplied by Quantachrome Corp.
The X-ray photoelectron spectroscopy experiments were car-ried out with a Kratos Axis Ultra Imaging X-ray photoelectronspectrometer equipped with a Mono(Al) anode and a multichanneldetector. Charge referencing was measured against adventitiouscarbon (C 1s, 284.8 eV). Shirley-type background was subtractedfrom the signals. The recorded spectra were treated with Gauss–Lorentz curves to determine the binding energy of the differentelement core levels.
values in microporous silica-based materials 1–8 (aluminum contents are determined
a due to NLDFT cumulative porevolume (cm3 g�1)
NLDFT cumulative pore volume due tomicropores (cm3 g�1)
Fig. 2. Cumulative pore volumes versus pore width for the samples with varyingnickel contents from top to bottom: 2(540), 1(540), 3(540) and 5(540).
80 V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86
The transmission electron microscopy experiments were per-formed at the Texas A&M University Microscopy and Imaging Cen-ter with a JEOL JEM2010 at a working voltage of 200 kV, with apoint resolution of 0.23 nm. All imaging magnifications were cali-brated using standards of SiC lattice fringes [23] for high magnifi-cations and commercial cross-line grating replica for lowmagnifications. The samples were dispersed in ethanol solution, asmall drop of the solution was transferred onto the top surface ofa 400-mesh carbon-film supported Cu grid which was previouslyglow discharged to achieve a better dispersion, and then the mate-rial was dried in air. The energy-dispersive spectroscopy (EDS)measurement was performed using an Oxford Instruments EDSdetector with INCA energy platform.
Magnetic measurements were performed on a Quantum DesignSQUID magnetometer MPMS-XL in an applied field of 1000 Oe inthe 2–300 K temperature range.
The nickel contents were obtained via an instrumental neutronactivation analysis performed at Texas A&M University Center forChemical Characterization and Analysis. The absolute concentra-tions of silicon and aluminum were determined via ICP and dupli-cated by AA analysis for randomly selected samples at AndersonAnalytical, Flynn, Texas, and varied within the range �35–40 wt%and 2–3 wt%, respectively.
3. Results and discussion
3.1. Pore structure
Table 1 lists the nickel content of the mixed oxides along withthe results of the BET N2 sorption–desorption measurements.EDS analysis performed on samples 2, 4 and 6 at five different posi-tions on each sample yielded the same Ni content per measure-ment within experimental error attesting to the uniformity of theNi distributions throughout the samples. Four of the eight iso-therms are shown in Fig. 1 and all eight are type-I similar to thefour shown. These curves are also similar to those obtained forthe Al2O3–SiO2–ZnO materials [8] and those for the Al2O3–SiO2–MnO system. It is seen that the bulk of the nitrogen has beensorbed at the lowest pressure which is 10�3P/P0. This is an indica-tion that the majority of the pores are in the micropore region. Poresizes as determined by the MP method indicate average pore sizesin the 8–13 Å range. For the DFT method to access pores belowabout 15 Å in diameter, data collection at pressures in the 10�7–10�3 range are required. However, the curves in Fig. 2 show thatat a pore width of 20 Å, the limit of microporosity, the cumulative
0
50
100
150
200
250
0.00 0.25 0.50 0.75 1.00Relative pressure (P/P0)
Volu
me
(cc/
g)
Fig. 1. Nitrogen sorption/desorption isotherms for the samples with varying nickelcontents from top to bottom: 2(540), 1(540), 3(540) and 5(540).
pore volumes vary from 85% for sample 3–65% for sample 2 of thetotal pore value as micropores.
It is evident that in each sample, irrespective of the metal oxideadded to the SiO2–Al2O3 the majority of nitrogen uptake occurswithin the pores below 20 Å, while the larger pores of 20–30 Åare responsible for the remaining nitrogen sorption. A more de-tailed characterization based on the data obtained at the P/P0 rangefrom 10�7 to 10�3 will be reported in a separate publication that isin progress.
3.2. X-ray and XPS data
The X-ray diffraction patterns of the materials are typical ofamorphous alumina–silica systems where only diffuse scatteringis observed. However, at nickel loadings of 13.4% and higher, theX-ray patterns show weak peaks at 2h = 37.13� (d = 2.42 Å) and43.24� (d = 2.09 Å) that can be assigned to NiO [24]. Nickel oxideparticles are also detected by the XPS experiments performed forthe materials with nickel concentrations of 13.4% and higher. Asseen from Fig. 3, the XPS pattern collected for material 5(450)exhibits a peak centered at 854 eV with a satellite at 860 eV [25].It should be emphasized that the observed lines are very wideand the presence of more than one type of nickel species cannotbe excluded. On the other hand, deconvolutions of such XPS pat-terns seem to be doubtful because two wide bands in Fig. 3 canbe fit to two peaks at 854.2 and 859.8 eV, belonging to ‘‘free” NiO,or to three peaks at 854.3, 857.8 and 859.9 eV, as a combination
840845850855860865870875880885890Binding Energy (eV)
Inte
nsity
Fig. 3. The XP Ni 2p3/2 spectrum of sample 5(540) calcined at 540 �C.
Fig. 5. The 27Al MAS NMR spectra recorded at a spinning rate of 6 kHz (from the topthe bottom: for sample 1(450) with a 27Al{1H} pulse sequence, for sample 1(450)with a CP H–Al pulse sequence and for sample 5(450) with a single-pulse sequence.
Fig. 6. The variable-temperature 2H NMR spectra recorded for 1-2H(450) (from thetop the bottom: 260 K, 220 K, 200 K and 150 K.
V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86 81
of free NiO and NiO interacting with the SiO2 support [26], or evenfour peaks at 854.3, 855.9, 858.3 and 860.6 eV. Thus, because of thenature of the materials investigated, the X-ray and XPS data areonly minimally instructive.
3.3. Solid-state NMR
Samples 1–8 calcined at 450 and 540 �C were characterized bysingle-pulse 29Si, 27Al and 1H MAS NMR spectra, and by the Hahn-echo 29Si MAS NMR spectra, recorded at different carrier frequen-cies and the T1 relaxation measurements performed for 1H, 2Hand 29Si nuclei [22]. The same technique was successfully appliedfor characterization of the systems doped by Mn2+ ions [7].
The NMR data were collected for non-calcined and calcined dia-magnetic samples 1 and 1(450). Their 29Si and 27Al MAS NMR spec-tra are typical of amorphous based systems [26]. According to the29Si CP and 29Si{1H} MAS NMR spectra of 1 and 1(450), site Q4(Si)remarkably dominates in the 29Si{1H} MAS NMR spectrum of1(450) (�110 ppm) while site Q3(Si) is observed as a weak andpoorly-resolved shoulder at �100 ppm (Fig. 4).
The 27Al MAS NMR spectrum of 1(450) shows an intense reso-nance at 53.5 ppm and a weak line at �0.7 ppm (Fig. 5). The rela-tive intensity of this weak line increases in the 27Al CP MAS NMRspectrum of 1(450).
Thus, the aluminum atoms are incorporated into the silica ma-trix as four-coordinated species [27] with the content of the extra-framework aluminum being small. All the organic components,clearly observed in the 13C CP MAS NMR spectrum of 1 (sharp res-onances at 169, 129, 60.3, 50.7, 33.3, 26.4 and 18.5 ppm), are notdetected in the calcined material 1(450). The 1H MAS NMR spec-trum of 1(450), exhibits a single resonance at 5.1 ppm, which canbe attributed to water molecules located within pores [28]. As evi-dence for this assignment it was found that the intensity of thissignal decreases for the sample packed into an NMR rotor immedi-ately after drying at 65 �C for 1 h. Also, the sample, treated withD2O and then dried at 65 �C (sample 1-2H(450)) shows a narrowresonance in the static 2H NMR spectrum, recorded at 260 K witha quadruple-echo pulse sequence (Fig. 6).
Thus quadrupolar interactions, typical for the solid state, arecompletely averaged due to fast isotropic or pseudo-isotropic mo-tions of the 2H nuclei (for example, rapid tetrahedral jumps) [29].Careful inspection of the 2H NMR spectrum did not reveal the pres-
Fig. 4. The 29Si MAS NMR spectra recorded at a spinning rate of 6 kHz (from the topto the bottom): for sample 1 with a 29Si{1H} pulse sequence, for sample 1(450) witha CP H–Si pulse sequence and for material 1(450) with a 29Si{1H} pulse sequence.
ence of a rigid powder-shaped line probably due to a fast exchangebetween ‘‘free” water molecules and molecules located directly onthe silica surface. In accord, the resonance line broadens to 2.2 and5.5 kHz when the temperature decreases to 230 and 220 K, respec-tively. At 200 K the line transforms towards a rigid deuterium pow-der pattern which is resolved at 150 K showing quadrupolesplitting of 148 kHz, typical of solid water (Fig. 6) [30]. It shouldbe emphasized that such temperature evolutions are typical ofsmall molecules located in pores of materials and having a highmobility [31].
The room-temperature 1H and 2H T1 NMR relaxation measure-ments performed for 1(450) and 1-2H(450) independently supportthe assignment. Generally spin-diffusion completely dominates inspin-lattice relaxation of protons in solids leading to an exponen-tial recovery of the magnetization [28,31]. However, the 1H spin-lattice relaxation in sample 1(450), spinning with a rate of 6 kHz,is non-exponential and described by a stretched exponential[32,33]:
I ¼ I0ð1� 2 expð�ðs=T1ÞbÞÞ ð1Þ
with 1H T1 = 0.22 s and b = 0.71. The same treatment of the non-exponential 2H T1 relaxation in sample 1-2H(450) results in the 2HT1 time and the b values of 0.014 and 0.76 s, respectively. Thenon-exponential behavior is typical of molecules situated in poresdue to the presence of T1 distributions [34], and the b parametersdetermined for the 1H and 2H relaxation curves are reasonably sim-ilar. Finally, the 1H, 2H T1 times of liquid water and water in pores
82 V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86
differ strongly. For example, the 2H T1 time in the bulk water is aslong as 0.42 s [34] versus 0.014 s measured in sample 1-2H(450).
The 29Si spin-lattice relaxation in diamagnetic sample 1(450) isalso non-exponential and well treated by a stretched exponentialfunction: 29Si T1 time = 55 s and b = 0.82. It should be noted thata very similar b parameter (0.77) was determined for aryl-sulfo-nate silica gels [35] and discussed in terms of spin-diffusion whenthe latter is limited [36,37]. In accounting for the spin-diffusioncoefficient of nuclei 29Si, which is equal to �10�19 m2 s�1 [37],spin-diffusion of 29Si nuclei could be actually effective at delaytimes s > 20 s in the inversion–recovery experiments. However,since the T1 measurements performed at spinning rates from 1.5to 11 kHz led to the identical (within 10% errors) 29Si T1 values,the spin-diffusion mechanism in 1(450) can be ruled out [31]. Gen-erally, the 29Si nuclei, for example, in silicate glasses, relax veryslowly with T1 times reaching >30 min in the absence of paramag-netic impurities [37]. Therefore, the shorter non-exponential 29Sirelaxation in sample 1(450) rather occurs by dipolar interactionsof nuclei 29Si with non-controlled paramagnetic centers, for exam-ple, O2, situated close to the surface of the silica matrix [38].
The 29Si MAS NMR spectra of samples 2(450)–8(450) or 2(540)–8(540) show the sideband patterns that are more pronounced atthe higher Ni loadings and invisible in the spectrum of the nick-el-free material (Fig. 7). The appearance of such intense sidebandpatterns has been earlier used as a criterion for invoking the incor-poration of paramagnetic metal ions into the silica matrix of thematerials [20,39].
According to another point of view, the appearance of the side-band patterns, themselves, even with shortened 29Si T1 relaxationtimes, shows the presence of paramagnetic centers only and doesnot prove the incorporation of paramagnetic metal ions [40]. The29Si MAS NMR experiments, performed for a sample, containingdiamagnetic material 1(450) carefully mixed with Ni(CH3-
COO)2 � 4H2O in an 1/1 weight ratio (marked through the text as1M), supported this idea. First, the 29Si MAS NMR spectrum of1M with 12% (weight) of Ni2+, recorded at a spinning rate of3 kHz (i.e. smaller than those in Fig. 7), shows the sideband patterninvisible in 1(450) at this rate. Second, the 29Si T1 time in 1M isremarkably shortened (0.29 s). It is obvious that these featurescannot be explained by quite close contacts between 29Si andNi2+ that are practically impossible in the mixture. On the other
Fig. 7. The 29Si MAS NMR spectra (from the top to the bottom) recorded for calcinedsamples 1(450), 2(450), 3(450), 4(450) (at 5 kHz) and 5(450), 6(450) and 8(450) (at6 kHz).
hand, the sideband and 29Si relaxation effects are still smaller thanthose observed for 4(450) with a significantly smaller nickelcontent.
As seen from the 29Si MAS NMR spectra in Fig. 7, the sidebandintensities increase sequentially from 0.0 to 13.4 Ni2+%, reach amaximum in material 5(450), do not change from 5(450) to6(450) and decrease again in 7(450) and 8(450). In spite of theabove-mentioned complexity in the spectral behavior of paramag-netic systems, this tendency shows clearly that in the materials,obtained at high nickel loadings, the content of paramagnetic cen-ters is becoming lower. This conclusion is valid even in the absenceof their more accurate localization.
Since the shape and the chemical shift of the 29Si isotropic res-onance in Fig. 7 do not change, the detected nuclei, belonging tothe silica lattice, are remote from paramagnetic centers. Closely-located 29Si nuclei, located in a first coordination sphere, for exam-ple, in moieties Si–O–Ni, incorporated into the silica matrix orsituated on its surface, could become ‘‘spectrally invisible” due tostrong paramagnetic effects [40]. Usually such nuclei can be ob-served as very wide lines in the Hahn-echo MAS NMR spectra, col-lected at different carrier frequencies and then summarized[14,40]. However, such Hahn-echo 29Si MAS NMR experimentswere unsuccessful and the spectra, collected in 1(450)–5(450) or1(540)–5(540), did not reveal even a remarkable loss in intensitiesof the 29Si resonances on going from diamagnetic 1(450) to para-magnetic systems. However, it is obvious that small contents ofsuch moieties cannot be completely excluded due to the well-known problem of NMR sensitivity particularly in the cases of rarenuclei with broadened signals.
The isotropic 27Al resonance in 2(450)–8(450) and 2(540)–8(540) is also accompanied by intense sidebands and observed at54 ppm (Fig. 5). The Hahn-echo 27Al MAS NMR experiments at dif-ferent carrier frequencies gave the same result. Again, the detectedAl atoms are in the silica framework and remote from paramag-netic centers. Finally, the 1H MAS NMR spectra of 2(450)–5(450)and 2(540)–5(540) exhibit sideband patterns with the isotropicresonances very close to that of the pore water in the diamagneticsample 1(450). In good agreement with increasing the Ni loadingsthe sideband effects become more pronounced from 2(450) to5(450) and the line-widths increase from 430 Hz in 1(450) to990, 1670, 4280 and 20,000 Hz in 2(450), 3(450), 4(450) and5(450), respectively. These strong 1H broadenings in combinationwith the relatively weak broadening effects observed for 29Si and27Al nuclei in Figs. 7 and 5, respectively, show accumulations ofthe paramagnetic Ni2+ centers within pores of the materials.
Additional data supporting this idea were obtained by NMRrelaxation experiments that revealed a very interesting effect.We have found that the 29Si T1 relaxation times are different inthe freshly-prepared paramagnetic samples, calcined at differenttemperatures (450 and 540 �C): the T1 times are remarkably short-er in the materials, calcined at 540 �C. For example, the 29Si T1 timein freshly-prepared 4(450) and 4(540) was measured as 1.0 and0.64 s, respectively. However, after one month, the values reduceand become identical: 0.24 and 0.23 s in 4(450) and 4(540), respec-tively. Then, after two and three months, the 29Si T1 values do notchange in the 10% limit. This phenomenon seems to be connectedwith the movement of the nickel centers. We believe that the rel-atively mobile water in systems 2(450)–8(450) and 2(540)–8(540)can promote slow room-temperature migrations of the paramag-netic centers along pores. The nickel migrations cause more homo-geneous redistributions through the volumes of the samples andincrease effectively a number of close contacts Ni���Si. The 29Siand 1H T1 relaxation times measured in such calcined paramag-netic systems are presented in Tables 2 and 3. As these data indi-cate, the proton relaxation is not exponential and described by astretched exponential with the b parameters between 0.65 and 0.7.
Table 2The 29Si MAS NMR T1 data collected for calcined materials 1(450)–8(450) and 1(540)–8(540) treated with stretched exponential functions (the data are obtained as aspinning rate of 6 kHz)
V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86 83
Since the 1H inversion–recovery experiments, performed atspinning rates 6–12 kHz, resulted in the identical 1H T1 times (forexample, 4.9 � 10�4–5.1 � 10�4 s in 8(450)), the 1H spin-diffusionmechanism is negligible [31] and the relaxation is controlled bydipolar interactions with paramagnetic centers [9,19,40]. The sameconclusion can be drawn for the 29Si T1 relaxation. It is also non-exponential (Fig. 8), well treated by a stretched exponentialfunction with the b values between 0.58 and 0.75 (Table 2) andis independent of spinning rates (it should be noted that in somecases the inversion–recovery curves can be treated in terms of a
Fig. 8. The 29Si inversion–recovery curve obtained at a spinning rate of 6 kHz forsample 4(450) treated with a stretched exponential (top) and bi-exponentialfunction (bottom).
bi-exponential process, for example, material 4(450) in Fig. 8);however, since such treatments give generally larger errors, theyare not discussed).
Finally, the linear correlation between the 1H and 29Si T1 relax-ation rates in Eq. (2)
R1ð1HÞ ðs�1Þ ¼ 1170R1ð29SiÞ ðs�1Þ � 720 ð2Þ
illustrates clearly the same relaxation mechanism operating forboth nuclei. This dipolar relaxation mechanism is expressed by
ð3Þ
where Np is the number of paramagnetic centers, se is the electronrelaxation time and the other constants are well known [41]. As fol-lows from Tables 2 and 3, the 29Si and 1H T1 times shorten stronglyeven at a Ni2+ concentration of 1.35%. Generally such effects areattributed to homogeneous distributions of paramagnetic centersthrough the volume of the samples [42] due to their incorporationinto the silica matrix. However, the 29Si isotropic resonances andthe their sidebands observed in material 4(450) or 4(540) haveshown the identical 29Si T1 times in contrast to the systems whereparamagnetic manganese centers were actually incorporated intothe silica lattice [21,22]. The remarkably shorter T1 relaxation timesof sidebands with respect to isotropic resonances have been also re-ported for nuclei 31P in systems with paramagnetic centers incorpo-rated into the matrix of the materials [43]. Thus, the relaxation datashow that the nickel centers are homogeneously distributed withinpores of our calcined systems in agreement with the TEM micro-graph illustrated for sample 5(450) in Fig. 9. It should be empha-sized that equal [29] Si T1 times are measured for the isotropicresonance and its sidebands in the mixture 1M where incorporationof Ni2+ is completely excluded.
As the data in Tables 2 and 3 reveal, a R1(1H)/R1(29Si) ratio, ob-tained for the 1H and 29Si relaxation rates, changes insignificantlyfrom one paramagnetic sample to another one and calculated as�900. At equal internuclear distances r(Si���Ni) and r(H���Ni), thedipolar mechanism must be more effective for 1H nuclei by a factorof 25 (see c2 in Eq. (3)). Thus, it is obvious that 1H nuclei are situ-ated closer to paramagnetic centers than 29Si nuclei by a factor of1.8. Again, the data correspond to accumulations of the paramag-netic centers within pores of the materials.
The formalism in Eq. (3) requires decreasing the 29Si T1 timesproportionally to [Ni]2. The pattern in Fig. 10 is more complex:the 29Si relaxation rate increases from 1(450) to 5(450), reaches amaximum and decreases at higher Ni contents. It should be notedthat this pattern correlates with the sideband effects observed in
Fig. 9. The TEM micrograph of calcined material 5(450).
Fig. 10. The 29Si spin-lattice relaxation rate as a function of the nickel loadings insamples 1(450)–8(450) (d); point (j) corresponds to the relaxation rate in 1M.
Table 4Magnetic properties of materials 2(450)–4(450) and 2(540)–4(540)
Fig. 11. Temperature dependence of the vT product for sample 4(450). The solidand dashed lines correspond to the best fit to Curie–Weiss law. Inset: Hysteresisloop for 4(450) at 1.8 K.
84 V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86
the 29Si MAS NMR spectra (Fig. 7). It is also important that the 29SiT1 time in mixture 1M obtained by a stretched-exponent treatmentat b = 0.64 is short (0.29 s) but the corresponding point in Fig. 10completely drops out of the plotted sequence.
As emphasized above, nuclei 29Si in materials 2(450)–8(450) re-lax via direct dipolar interactions with the paramagnetic centers(Eq. (3)). However, the linear dependence of the 29Si relaxation rateon the Ni concentration from 1(450) to 4(450) plotted on a loga-rithmic scale shows a slope of 0.6 instead of 2 expected theoreti-cally (see Eq. (3)). This disagreement could be explained byaggregations of paramagnetic centers and reducing the electronrelaxation time, se, due to exchange interactions [20]. The devia-tion of material 5(450) from the linearity in Fig. 10 is too large,however, and the further decrease in the relaxation rate rather cor-responds to a partial transformation of the paramagnetic nickelcenters into a non-paramagnetic state, for example, Ni0 [44]. Then,extrapolation of the linear section into the region of 5(450) pro-vides an estimation of the content of the metal nickel in this mate-rial as �50%. As described below, even this rough estimation agreeswell with magnetic measurements performed on powder samples2(450)–5(450) and 2(540)–5(540) to reveal the nature of the para-magnetic centers.
3.4. Magnetic measurements
The magnetic behavior of the materials, calcined at 450 and540 �C, are similar. Samples 2(450) and 2(540) exhibit room-temperature vT values of 0.254 and 0.263 emu g�1 K, respectively(Table 4).
Taking into account the elemental analyses, the spin-state ofthe magnetic centers is estimated as S = 1, g = 2.33 and g = 2.37 typ-ical of the magnetically-isolated Ni2+ ions (S = 1, g = 2.0–2.8) [45].Below 10 K, the vT value decreases due to zero field splitting ef-fects [46] and the temperature dependence of the magnetic sus-ceptibility can be fit to the Curie–Weiss law giving negativeWeiss constants, h, of �0.8 K and �1.3 K for the materials 2(450)and 2(540), respectively. Similar room-temperature magneticproperties are observed in 3(450) and 3(540) (Table 4). However,upon cooling, the vT values decrease indicating random antiferro-
magnetic interactions between magnetic centers. In accord withthis assumption is the fact that the temperature dependences arewell described by the Curie–Weiss law leading to a negative Weissconstant h of �2.2 K and �3.2 K for samples 3(450) and 3(540).The room-temperature vT values in samples 4(450) and 4(540)are calculated as 0.96 and 0.84 emu g�1 K, respectively. In this case,however, the vT magnitude gradually increases to a broad maxi-mum at �10 K and then it decreases again (Fig. 11). A treatmentin terms of the Curie–Weiss law leads to positive Weiss constantsh of +11.5 and +9.7 K for 4(450) and 4(540), respectively. Althoughbulk NiO is an antiferromagnet below TN = 525 K [47], this resultsuggests ferromagnetism due to the formation of NiO nanoparticles[48]. In fact, according to Néel [49], fine particles of an antiferro-magnetic material should exhibit magnetic properties such assuperparamagnetism and weak ferromagnetism, wherein the per-manent magnetic moment is attributed to an uncompensatednumber of spins on two sublattices.
The zero field cooling–field cooling (ZFC-FC) measurementsperformed in 4(450) and 4(540) at 10 Oe confirm this conclusionand show a 3D magnetic ordering at 6 K. This ordering is accompa-nied by a history dependence of the magnetization process in thatthe remnant magnetization does not follow the FC-curve and cor-responds to glassy behavior [50]. The hysteresis observed at 1.8 Kwith coercivity at �750 Oe and 400 Oe for the materials 4(450)and 4(540) is depicted in Fig. 11 (inset) by the region ±7500 Oe. In-deed, NiO nanoparticles were found to exhibit large coercive fieldsat low temperature, due to surface anisotropy [51–53]. Finally, thefrequency dependence of the AC magnetic susceptibility studiedbelow the phase transition (Fig. 12) agrees with the presence of adegree of cluster-glass like behavior with the Mydosh parameter,/ ¼ DTm=Tg=D log x [54], being estimated as 0.1. Here DTm is theshift of the peak in v0, log x is the logarithm of the applied fre-quency, and Tg is freezing temperature.
0.6
0.7
0.8
0.9
1
1.1
1.2
1.3
1.4
2 6 10 12Temperature (K)
χ· 1
03 (em
u/g)
0
0.05
0.1
· 103 (e
mu/
g)
1 Hz10 Hz100 Hz1 kHz
χ
4 8
Fig. 14. Temperature dependence of the real v0 (full symbols) and imaginary v00
(open symbols) components of the AC susceptibility with oscillating field of 6 Oe atdifferent frequencies for 5(450).
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
2 6 10Temperature (K)
· 103 (e
mu/
g)
0
0.05
0.1
0.15
0.2
· 103 (e
mu/
g)
1 Hz10 Hz100 Hz1 kHz
χ
4 8
χ
Fig. 12. Temperature dependence of the real v0 (full symbols) and imaginary v00
(open symbols) components of the AC susceptibility with oscillating field of 3 Oe atdifferent frequencies for 4(450).
V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86 85
The variable-temperature magnetic measurements in samples5(450) and 5(540) (corresponding to the top point of the relaxationdependence in Fig. 10) differ completely from those in materialswith lower nickel concentrations. These data show a substantialtemperature-independent component of the ferromagnetic natureusually observed for particles Ni0 [53]. The room-temperaturefield-dependent magnetization in this material reveals an unusu-ally fast rise at low field without saturation even at 70 kOe(Fig. 13).
Generally such behavior is attributed to the formation of ferro-magnetic Ni0 nanoparticles [55] which have blocking temperatureclose to the ordering temperature of bulk nickel Tc = 630 K [56].Nevertheless, the curves do not follow the simple Langevin expres-sion valid for a superparamagnet [57]. Since the pattern M versus Hhas a nonzero slope even at higher H, the curves were fitted to themodified Langevin function [58–60]:
M ¼ MsLðlpH=kBTÞ þ vaH ð4Þ
where lp is the magnetic moment of a nanoparticle, kB is the Boltz-mann constant, L(x) = cothx � 1/x, and va is the high-field suscepti-bility. The plot of magnetization versus field at T = 300 K shown inFig. 13 reveals superparamagnetism with lp = 14,000lB, and Ms =1.4 emu g�1. Using the analytic data and assuming a magneticbehavior of the particle close to that of bulk metal and a saturation
Fig. 13. Room-temperature field-dependent magnetization curve for 5(450). Thesolid line corresponds to the best fit to Eq. (4). Inset: Hysteresis loop for 5(450) at300 K.
magnetization of 55 emu g�1 [61,62] one can estimate the Ni2+ toNi0 transformation as 30%. In order to understand better the abovefeature better, the lp magnitudes were determined from the equa-tion lp/lB = MsqV, where q and V are the density and volume of theparticle, respectively. Assuming for bulk nickel a spherical particlewith q = 8.9 g cm�3 and Ms = 55 emu g�1 gives a diameter D of8.1 nm. Materials 5(450) and 5(540), which exhibit similar mag-netic behavior, also show a narrow hysteresis at 300 K with coerciv-ity of approximately 100 Oe (Fig. 13, inset), again in accord withpreviously reported data for Ni nanoparticles [63]. The frequencydependence of the AC magnetic susceptibility reveals two peaks(Fig. 14) corresponding to a two phase system.
The Mydosh parameter, /, was estimated as 0.14 in a full agree-ment with the superparamagnetic behavior. Finally materials6(450)–8(450) with higher nickel concentrations exhibit mostlyroom-temperature ferromagnetic properties.
As follows from the collective data, the magnetization and 29Si T1
relaxation experiments carried out for 5(450) are in good agree-ment, and nanoparticles with sizes similar to those calculated bythe magnetic measurements can be found in the TEM micrograph(Fig. 9). In contrast, standard deconvolution procedures, performedfor the wide XPS bands in Fig. 3, did not reveal the presence of peaksat 852.3 and 868.0 eV for nickel metal [26]. This result is a goodillustration of the intrinsic problem of the XPS method which probesonly near-surface atom concentrations. The Ni0 cluster sizes are‘‘invisible” in most of the X-ray diffraction patterns of paramagneticmaterials even with high nickel loadings. However, fortunately,
0
20
40
60
80
100
120
0 10 20 30 40 50 60 702Θ
Inte
nsity
Fig. 15. The X-ray pattern of material 5(450).
86 V.I. Bakhmutov et al. / Microporous and Mesoporous Materials 118 (2009) 78–86
these Ni0 clusters are well observed in the X-ray pattern recordedfor material 5(450) (Fig. 15).
Partial reduction of Ni2+ to Ni0, responsible for the magneticproperties of materials 5(450)–8(450) and 5(540)–8(540), is notsurprising and is attributed to the presence of organic templatesused in the synthetic procedures. It is interesting, however, thatsample 8(450), heated in oxygen at 540 �C for 6 h, exhibits the same29Si NMR MAS spectrum and the same 29Si T1 time (0.33 s atb = 0.73). It is generally accepted that the metallic nickel is com-pletely oxidized in an oxygen atmosphere at 430 �C [64,65]. In thisconnection we believe that the above stability is explained by thelocation of the nickel particles within a NiO shell.
4. Conclusions
A series of materials SiO2–Al2O3–NiO with a predominant poresize in the range 8–20 Å have been synthesized by the sol–gelmethod where Ni2+ loadings have been varied between 0 and30 wt%. The materials calcined at 450 and 540 �C were studiedby X-ray diffraction, X-ray photoelectron spectroscopy, transmis-sion electron microscopy, variable-temperature magnetic suscepti-bility measurements and 29Si, 27Al, 1H, 13C solid-state MAS NMRspectra recorded with single-pulse and Hahn-echo pulse se-quences. It has been shown that solid-state NMR and 1H and 29Sispin-lattice NMR relaxation in such amorphous paramagnetic sol-ids can be successfully used for their characterizations. The 1H and29Si spin-lattice NMR relaxation is a non-exponential process andgoverned by direct dipolar interactions with paramagnetic centerswhile the spin-diffusion mechanism is negligible. The 29Si and 1Hrelaxation rates, obtained in the limits of the stretched exponen-tial, nicely reflect the distribution of paramagnetic centers. It wasfound that aluminum atoms as four-coordinated species are incor-porated into the silica matrix while nickel centers are accumulatedwithin pores. The nature of the nickel centers was probed by var-iable-temperature magnetic susceptibility measurements sup-ported by XPS and X-ray experiments. The Ni species areconcluded to be NiO and Ni0 (observed at high nickel loadings)aggregated into nanoparticles. The Ni0 nanoparticles are responsi-ble for the room-temperature ferromagnetic behavior of thematerials prepared with high nickel loadings whereas the low-temperature ferromagnetism, observed in the materials with lowernickel loadings, is caused by NiO nanoparticles.
Acknowledgments
The work reported here was supported by the National ScienceFoundation under Grant Nos. DMR-0332453 and CHE-0234931, forwhich grateful acknowledgment is made. K.R.D. is grateful to theDepartment of Energy and the Welch Foundation for financial sup-port. The SQUID magnetometer was purchased with a grant fromthe National Science Foundation. We sincerely thank Dr. W.D.James and Mr. M. Raulerson for performing the neutron activationanalysis on our samples.
(2006) 81.[9] M.J. Duer, Solid-State NMR Spectroscopy: Principles and Applications,
Blackwell Science, Oxford, 2002.[10] C. Bonhomme, C. Coelho, N. Baccile, C. Gervias, T. Avaiz, F. Babonneau, Acc.
Chem. Res. 40 (2007) 738.[11] G. Mali, A. Ristic, V. Kaucic, J. Phys. Chem. B 109 (2005) 10711.[12] A. Kubo, T.P. Spaniol, T. Terao, J. Magn. Reson. 133 (1998) 330.[13] R. Rosas-Salas, J.M. Dominguez, F. Rachdi, C. Alvarez, Stud. Surf. Sci. Catal. 146
(2003) 131.[14] L. Canesson, A. Tuel, Chem. Commun. (1997) 241.[15] H. Hiese, F.H. Kohler, X. Xie, J. Magn. Reson. 150 (2001) 198.[16] A.N. Clayton, C.M. Dobson, C.P. Grey, J. Chem. Soc. Chem Commun. (1990) 72.[17] C.P. Grey, C.M. Dobson, A.K. Cheetham, R.J.B. Jakeman, J. Am. Chem. Soc. 111
2005, pp. 179–181.[20] D. Goldfarb, Zeolites 9 (1989) 509.[21] V.I. Bakhmutov, B.G. Shpeizer, A. Clearfield, Magn. Reson. Chem. 44 (2006) 985.[22] V.I. Bakhmutov, B.G. Shpeizer, A. Clearfield, Magn. Reson. Chem. 45 (2007) 118.[23] Z.P. Luo, Acta Mater. 54 (2006) 47.[24] T. Ahmad, K.V. Ramanujachary, S.E. Lofland, A.K. Gaguli, Solid State Sci. 8
(2006) 425.[25] G. Poncelet, M.A. Centeno, R. Molina, Appl. Catal. A Gen. 288 (2005) 232.[26] R. Mokaya, Micropor. Mesopor. Mater. 44–45 (2001) 119.[27] S. Ganapathy, K.U. Gore, R. Kumar, J.P. Amoureax, Solid State Nucl. Magn.
Reson. 24 (2003) 184.[28] X. Xue, M. Kanzaki, Solid State Nucl. Magn. Reson. 31 (2007) 10.[29] R.J. Witterbort, M.G. Usha, D.J. Ruben, D.E. Wemmer, A. Pines, J. Am. Chem. Soc.
110 (1998) 5668.[30] A.J. Benesi, A.M.W. Grutzeck, B. O’Hare, J.W. Phair, J. Phys. Chem. B 108 (2004)
17783.[31] S. Hayashi, Solid State Magn. Reson. 3 (1994) 323.[32] G. Williams, D.C. Watts, Trans. Faraday Soc. 66 (1970) 80.[33] C.A. Angel, L.M. Torell, J. Chem. Phys. 78 (1983) 937.[34] R. Chen, P.E. Stallworth, S.G. Greenbaum, R.L. Kleinberg, J. Magn. Reson. A 110
(1994) 77.[35] M.H. Alaimo, J.E. Roberts, Solid State Nucl. Magn. Reson. 8 (1997) 241.[36] B. Maiti, B.R. McGarvey, J. Magn. Reson. 58 (1984) 37.[37] M.G. Motruza, R. Dupree, D. Holland, J. Non-Cryst. Solids 281 (2001) 108.[38] C.A. Fyfe, D.H. Brouwer, J. Am. Chem. Soc. 126 (2004) 1306.[39] C. Montes, M.E. Davis, B. Murray, M. Narayana, J. Phys. Chem. 94 (1990) 6425.[40] L. Canesson, Y. Boudeville, A. Tuel, J. Am. Chem. Soc. 119 (1997) 10754.[41] G. Mali, V. Kaucic, Solid State Nucl. Magn. Reson. 12 (1998) 243.[42] S.H. Chen, S.P. Sheu, K.K. Chao, J. Chem, Soc. Chem. Commun. (1992) 1504.[43] B. Suttler, R.E. Taylor, L.R. Hossner, D.W. Ming, Soil Sci. Soc. Am. J. 66 (2002)
5.[45] E.A. Bourdreaux, L.N. Mulay, Theory and Application of Molecular Magnetism,
Wiley, New York, 1976. p. 210.[46] O. Kahn, Molecular Magnetism, VCH Publisher, New York, 1993. p. 26.[47] C. Kittel, Introduction to Solid State Physics p. 448, seventh ed., Wiley and
Sons, New York, 1996 (Chapter 15).[48] R.H. Kodama, A. Salah, A.E. Berkowitz, Phys. Rev. Lett. 79 (1997) 13934.[49] L. Neel, in: C. Dewitt, B. Dreyfus, P.D. Gennes (Eds.), Low Temperature Physics,
Gordon and Beach, New York, 1962, p. 413.[50] R.L. Carlin, Magnetochemistry, Springer-Verlag, Berlin, Heidelberg, 1986.[51] L. Li, L. Chen, R. Qihe, G. Lia, Appl. Phys. Lett. 89 (2006) 134102.[52] E. Winkler, R.D. Zysler, M. Vasquez Mansilla, D. Fiorani, Phys. Rev. B 72 (2005)
132409.[53] Y. Ichiyanagi, N. Wakabayashi, J. Yamazaki, S. Yamada, Y. Kimishima, E.
Komatsu, H. Tajima, Physica B 329–333 (2003) 862.[54] J.A. Mydosh, Spin Glasses, Taylor and Francis, London, 1993.[55] C. Parada, E. Moraan, Chem. Mater. 18 (2006) 2719.[56] H. Amekura, Y. Fudamoto, Y. Takeda, N. Kishimoto, Phys. Rev. B 71 (2005)
172404.[57] P. Dutta, A. Manivannan, M.S. Seehra, N. Shah, G.P. Huffman, Phys. Rev. B 70
(2004) 174428.[58] A. Punnoose, T. Phanthavady, M.S. Seehra, N. Shah, G.P. Huffman, Phys. Rev. B
69 (2004) 054425.[59] S.A. Makhlouf, F.T. Parker, A.E. Berkowitz, Phys. Rev. B 55 (R14) (1997) 717.[60] M.S. Seehra, P. Dutta, H. Shim, A. Manivannan, Solid State Commun. 129 (2004)
721.[61] C. Estournes, T. Lutz, J. Happich, T. Quaranta, P. Wissler, J.L. Guille, J. Magn.
Magn. Mater. 173 (1997) 92.[62] Y. Wang, Q. Zhu, H. Zhang, J. Mater. Chem. 16 (2006) 1212.[63] J.H. Hwang, V.P. Dravid, M.H. Teng, J.J. Host, B.R. Elliott, D.L. Johnson, T.O.
Mason, J. Mater. Res. 12 (1997) 1076.[64] C. Hoang-Van, Y. Kachaya, S.J. Teichner, Y. Arnaud, J.A. Dalmon, Appl. Catal. 46
(1989) 281.[65] S. Narayan, R.P. Unnikrishan, V. Vishwanathan, Appl. Catal. 129 (1995) 9.
