Characterization of Cuban and Brazilian natural zeolites by photoacoustic spectroscopy and electron paramagnetic resonance

: This report describes the photoacoustic and electron paramagnetic resonance investigations of Brazilian and Cuban zeolites. Photoacoustic optical absorption measurements indicate the presence of iron (Fe 3+ ) ions with their respective transition bands for both zeolites. Two species of manganese (Mn 2+ and Mn 3+ ) were identified in the Cuban sample and the electronic transitions assigned. Iron and manganese ions were confirmed through nonradiative relaxation ( τ ) and characteristic diffusion ( τ β ) times evaluation, whose values were found to be τ BRA = 5.40 ms, τ CUB = 4.60 ms, τ β BRA = 387 μ s and τ β CUB = 305 μ s. Crystal field (Dq-BRA /Dq-CUB = 1048 cm – 1 /945 cm – 1 ) plus Racah (B-BRA /B-CUB = 457 cm – 1 /813 cm – 1 and C-BRA /C-CUB = 3655 cm – 1 /2496 cm – 1 ) parameters were assessed as well. Paramagnetic resonance corroborated Fe 3+ ions present in the Brazilian zeolite occupying sites showing axial and/or rhombic symmetry distortions. For the Cuban sample, results reveal the characteristic hyperfine sextet lines of Mn 2+ overlapping the Fe 3+ line. Values of Landé factor and isotropic hyperfine splitting constant were found to be 2.0 and 9.7 mT, respectively. This tells us that the Mn 2+ lies in octahedral symmetry probably replacing calcium ions and point towards an ionic bonding character of the Mn 2+ and its surroundings.


INTRODUCTION
Zeolites can be both natural or synthetic and their definition, strictly speaking, embraces only hydrated crystalline aluminosilicates formed by SiO 4 and AlO 4 tetrahedrons linked by oxygen atoms.These structures contain open vessels in the form of channels with diameter ranging from 4.2 Å to 7.4 Å and cages.Zeolites frameworks could accommodate guests (e.g.alkaline metals) in either interstitial or substitutional sites, compensating their negative charges.Therefore, due to the easy exchange of cations at low temperatures, zeolites have been mainly employed as catalyst for petroleum cracking and synthesis of organic compounds.Owing to their porous structures, zeolites have also been used in gas and water/wastewater cleaning processes Breck (1974), Dyer (1988), Luna & Schuchardt (2001), Krstić et al. (2018).
By means of photoacoustic spectroscopy (PAS), it is possible to detect the absorption spectra of some of the zeolites' constituents; here, iron (Fe 3+ ) and manganese Mn 2+ and Mn 3+ ions.The PA signal is generated due to the absorption of modulated or pulsed light by absorbing centers followed by nonradiative decay (localized and periodic heating) Vargas & Miranda (1988), Almond & Patel (1996).Absorbing centers are distributed throughout the sample, and the heat generated at their locations diffuses toward the sample surface and there gives rise to the acoustic signal, which is detected by a microphone.When there is a time lag between the photoacoustic signals generated by the sample's constituents, the individual spectrum of each constituent can be distinguished using the phase-resolved PAS (PRPAS).Through electron paramagnetic resonance (EPR), the identification and status of the sites of Fe 3+ and Mn 2+ ions are feasible.
Particularly, Brazilian and Cuban zeolites have been investigated from the point of view of chemical composition (wt%), textural analysis (specific area, pore volume and mean pore diameter), TPD (temperature programmed NH 3 desorption), nitrogen content by Kjeldahl method, structural (x-ray diffraction accompanied by Rietveld refinement), and micro-structural (transmission electron microscopy) characterizations Baptista-Filho et al. (2011), Monte et al. (2009), Baptista-Filho (2011), Riter (2011).Although many properties of natural zeolites have been found hitherto in the literature, as far as the authors know, there are no information about PA assessment of optical transition bands, crystal field and Racah parameters, beyond the determination of nonradiative relaxation and characteristic diffusion times of iron and manganese ions within the zeolite framework.Therefore, this work is intended to cover such lack of data and contribute to the zeolite community providing new results.

THEORY OF PHASE-RESOLVED METHOD (PRM)
Phase-resolved photoacoustic spectroscopy (PRPAS) was applied to separate the spectra generated by different chemical constituents of the sample.The separation is accomplished when the absorbing centers, e.g., A and B, have different nonradiative relaxation time τ or/and characteristic diffusion time Baesso et al. (1989), Corrêa et al. (2011), what is closely dependent on how these absorbing centers are depth-distributed into the sample.Thus, a finite time lag between the photoacoustic signals generated by the chemical constituents becomes measurable using a phase-sensitive detection technique.To simplify the understanding of the phase-resolved method, a vector diagram is depicted in Figure 1, where each vector symbolizes the signal intensity at a specific wavelength (λ A and λ B ) absorbed preferably by centers A and B, respectively.The resulting signal (R) is the sum of vectors A and B (signals related to the absorbing centers here identified as A and B).The angles θ A and θ B are respectively the phases between the signals A and B with respect to an external reference.X (in-phase component of the signal) and Y (quadrature component of the signal) are the projections of the vector R on the X-and Y-axis, respectively.To resolve the spectrum in B after registering the signal with the change of the wavelength, the task consists of finding a new reference axes for which the projection of the vector A, for instance, on the new X-axis (or Y-axis) is null.Such procedure might be achieved mathematically by rotating the original reference axes by an angle ϕ in such a way that the projection of the vector B, i.e. (X ϕ ), is just equal to the projection of the vector R, that is, totally free from the contribution of the projection of A (Y ϕ ).At a given angle ϕ, this condition is achieved when: (1) Whenever the condition above is satisfied, the phase between the signals A and the external reference can be determined.As the X ϕ component has no contribution of A, the angle between A and the new X-axis is 90 • .Thus, as shown in the vector diagram (Figure 1), ϕ + θ A = 90 • César et al. (1984, 1985), Pessoa-Júnior (1985).
Lock-in amplifiers are based on phase-sensitive detection technique to single out the components of the signal at a specific reference frequency and phase.X and Y represent the in-phase and quadrature components of a signal, respectively, and they are measured directly by the lock-in amplifier.Thereby a spectrum proportional to that generated by one of the absorbing centers is computed through the Equation 1 at a given phase ϕ.Nowadays, lock-in amplifier has also the possibility to add a shift ϕ in the phase of the signal and thus record directly the absorption spectrum of a unique chemical specie.

Samples
Powder samples were provided by the Brazilian Agricultural Research Corporation (EMBRAPA): Cuban natural zeolite (clinoptilolite, furnished by "Celta Brasil", Cotia, SP, Brazil) and a Brazilian sedimentary zeolite from a natural occurring mixture of clays, zeolite and quartz extracted from central-northern Brazil Monte et al. (2009).For the sake of simplicity, Cuban and Brazilian zeolites will be labeled as CUB and BRA, respectively.

Photoacoustic spectroscopy (PAS) measurements
PA experiments were performed exciting the samples by a modulated monochromatic light after passing the collimated beam of 0.6 kW xenon arc lamp (Oriel Corporation) through a monochromator (Oriel-model 77250).A step-motor was employed to change the incident angle of the radiation beam on the diffraction grating of the monochromator so that the PA spectra were registered scanning the wavelength from 300 nm to 750 nm.The light was modulated at 17 Hz by a mechanical chopper (Stanford Research Corporation-SR540) that was driven by the sine output signal from the function generator of the lock-in amplifier (Stanford Research Corporation-SR830).The modulation frequency of the chopper was used as external reference.The PA signal generated by a 2 g powder sample placed inside a conventional PA cell (MTEC Photoacoustics, Inc., Ames, IA, USA) was detected by a condenser microphone located inside the PA cell.The time constant of the lock-in was set at 300 ms and a sampling rate of 1 Hz was applied.To take into account the dependence of the intensity of the monochromatic light on the wavelength, both X and Y components of the PA spectra were normalized by the X and Y components of a black-body sample.Nonradiative relaxation time τ and characteristic diffusion time τ β were also determined by varying the modulation frequency of the light source between 20 Hz and 60 Hz at a fixed wavelength.

Electron paramagnetic resonance (EPR) measurements
The EPR measurements were performed at room temperature with a Brucker (Elexys model E500) spectrometer operating at microwave X-and Q-bands.A conventional BRUKER strong pitch (fiberglass, carbon, KCl) standard (g = 2.0028) was used to calibrate the external magnetic field.The microwave frequency is precisely measured (δf = 10 -4 GHz) by the spectrometer itself.

RESULTS AND DISCUSSION
Figure 2 shows the PA spectrum of the BRA zeolite.The PRM applied for three different ϕ values is depicted in Figure 2a, revealing any change in the spectrum profiles.Such behavior is typical for samples with only one absorption center.Gaussian curves representing the optical absorption bands attributed to the electronic transitions of Fe 3+ ions Manhães et al. ( 2002), Mota et al. (2009) were combined to fit the data (Figure 2b).These transitions are assigned to 2b) with absorption bands centered respectively at 413 nm (24231 cm -1 , Hannoyer et al. ( 1992)), 522 nm (19157 cm -1 , Gargori et al. ( 2017)), 568 nm (17606 cm -1 , Karickhoff (1973)) and 645 nm (15504 cm -1 , Manhães et al. ( 2002)).Do data in the literature indicate that bands b 2 and b 3 here identified are attributed to the same transition, that is, According to the ligand field theory, these transitions are expected for Fe 3+ ions located in octahedral or tetrahedral symmetry Sugano et al. (1970) sites.One point which corroborates such assertion is the fact that the BRA zeolite is a natural material whose composition includes, among others, clay minerals Monte et al. ( 2009); besides, it is well established the presence of iron ions into the Brazilian soil composition Manhães et al. ( 2002), Mota et al. (2008Mota et al. ( , 2009)), Baptista-Filho et al. (2011).
Figure 3a shows the measured CUB zeolite spectrum (solid line) and the effect of the PRM using ϕ = 250 • , 120 • , and 90 • values.For these angles, structural changes were observed on the experimental spectrum profile.Contributions of Fe 3+ to the optical spectra are resolved for ϕ = 250 • (Figure 3b).Again, four Gaussian curves were used to fit the experimental data and at least three electronic transitions of Fe +3 could be identified at 380 nm (26316 cm -1 ), 485 nm (20619 cm -1 ), and 547 nm (18282 cm -1 ).The electronic transitions Hannoyer et al. (1992), Karickhoff (1973) 3b).For CUB zeolite, these electronic transition are centered at lower wavelength values compared to BRA zeolite.Such shift could be probably due to a different electronic screening experienced by Fe 3+ ions within uneven zeolitic matrices.
At ϕ = 90 • , four Gaussian curves centered at 350 nm, 415 nm, 491 nm, and 694 nm were combined to fit the experimental data.According to the literature, the first two bands are related to Mn 2+ and the third band is associated with Mn 3+ .Characterizing Mn 2+ in zinc phosphate glass, Ravikumar Ravikumar et al. (2003) associated the bands at 350 nm and 415 nm (labeled as c 4 and c 5 in Figure 3c) as being due to the 6 A 1g (S) → 4 T 2g (D) and 6 A 1g (S) → 4 A 1g (G) electronic transitions, respectively.Whilst Baesso and co-workers investigating soda-lime glasses, have found a broad band centered at ≈ 491 nm (labeled as c 6 ).They attributed this band to Mn 3+ ions Baesso et al. (1989).
By means of PAS, the nonradiative relaxation time τ and the characteristic diffusion time τ β were determined.Figure 4a shows the PA signal amplitude for the Cuban zeolite as a function of modulation frequency f.A linear fit revealed that the PA signal depended on frequency, as f -1 .For thermally thick samples, that is, samples whose thicknesses are greater than the thermal diffusion length, either thermoelastic bending or thermal dilatation result in the same f -1 frequency dependence on the signal amplitude Filho et al. (2009).Nonetheless, the latter discloses a constant PA phase as a function of modulation frequency.This PA phase behavior was never observed in our measurements.Thereby, the thermoelastic sample bending is the prevailing mechanism for the PA signal generation Perondi & Miranda (1987), Rousset et al. (1983).The reason behind our assertion is shown in Figure 4b, where the best fit for the PA phase is depicted as a function of modulation frequency for the Cuban zeolite according to Equation 2 Filho et al. ( 2009), Mota et al. (2010), leaving τ and τ β to be adjusted.
where ω = 2πf.We refer to the works of César and Pessoa-Júnior (César et al. 1984, 1985, Pessoa-Júnior 1985) for a more detailed discussion on a modified Rosencwaig and Gersho theory.described by Kumar et al. (1996) and Pedro et al. (2009), the crystal field parameter (Dq) and the Racah interelectronic repulsion parameters (B and C) have been evaluated.Doing so, values of which gave us the best fit with observed data were found to be Dq = 1048 cm -1 , B = 457 cm -1 and C = 3655 cm -1 for iron ions in BRA-zeolite and Dq = 945 cm -1 , B = 813 cm -1 and C = 2496 cm -1 for iron ions in CUB-zeolite.The energy values used to assess these parameters were 24213 cm -1 , 19157 cm -1 and 15504 cm -1 for the BRA sample and 26316 cm -1 , 20619 cm -1 and 18282 cm -1 for the CUB sample.Aluminosilicate raw materials, as the name suggests, are composed of superimposed layers of silica/alumina-rich tetrahedrons and octahedrons Cairns-Smith & Hartman (1986), Partheniades (2009).Iron and manganese ions encountered through their skeleton might be as interstitial or substitutional impurities so that, if the Fe 3+ replaces an Al 3+ in an octahedral sheet, the balance of charges remains constant and no distortion is to be expected (isomorphous replacement) Joshi et al. (1993).However, if the Fe 3+ substitutes a Si 4+ in a tetrahedral site, some degree of distortion may be found due to the resulting unbalanced charges Meunier (2005).Concerning the Mn 2+ ions, they can possibly be replacing a Ca 2+ or Mg 2+ ion into the mineral matrix Franco & Rossi (2003), Gunasekaran & Anbalagan (2008).Thus, to confirm that the chemical species Fe 3+ and Mn 2+ are present in Brazilian and Cuban zeolites, X-and Q-band EPR spectra from both samples were taken.For the ferric ion, when it is in a substitutional tetrahedrally coordinate original Si 4+ position, the unbalanced charge yields low symmetry of the electric field of the Fe 3+ environment.In this case, the spin Hamiltonian which represents the high-spin Fe 3+ , (d5, S=5/2), includes the Zeeman term and the zero field splitting (ZFS) one: where β is the Bohr magneton, D and E are the axial and the rhombic symmetry distortions parameters Mansanares et al. (1989), Goldfarb et al. (1994).
Q-band measurements show better resolved spectra, thus allowing better identification of the present species; hence, only the Q-band are being presented.Figure 5 presents the measured Qband EPR spectrum for the BRA-zeolite and lines are attributed to Fe 3+ ions, since Fe 2+ ions are not usually observed at room temperature and, although some rare earth paramagnetic elements have been detected in some zeolites Rizo & Peraza (1997), their concentrations are very small compared to those of the elements iron and manganese.Eventual rare earth resonance lines will be immersed in the much more intense lines of iron and manganese ions in our spectra.Moreover, rare earth EPR lines are generally detected at low temperatures and our experiments were performed at room temperature.No other paramagnetic species are supposed to be present.The experimental spectra were fitted by a sum of n first derivatives of the Lorentzian equation Mota et al. (2009): where H is the applied external magnetic field, A i is the peak-to-peak signal amplitude, H 0i is the magnetic resonance field and ∆H i is the peak-to-peak line width and n is the adequate number of equations in each case of overlapped lines.
Looking at Figure 5a, the more intense line in the center of the spectrum was fitted by Equation 4considering n = 2, yielding two lines with g values of 2.11 (curve 1) and 2.03 (curve 2), attributed to Fe 3+ ions in nondistorted octahedral environment Mota et al. (2009), Sengupta et al. (2010) either in interstitial zeolite network sites or in clusters structures due to strong interactions between iron ions Goldfarb et al. (1994), Soulayman et al. (2004).These are the main contributions to the spectrum.Figure 5b shows the low field spectrum region (left side of Figure 5a from 0 kG to 10 kG).The presence of a weak line distribution at g values of 4.9, 3.3, 2.8, 2.6 and 2.4 (arrows from 3 to 7, respectively) was identified.For the sake of visualization, at ≈ 11.5 kG arrow 8 indicates the sixth line, g = 2.12, belonging to the aforementioned distribution (it is displayed in Figure 5a and not in Figure 5b just because of scale purposes).They correspond to Fe 3+ ions in tetrahedral sites with strong rhombic distortion Mansanares et al. (1989), Berger et al. (1995).Diagonalizing H, Equation 3, in the S = 5/2 manifold we find three doublets whose energies depend upon the D/E relation.Thus, one obtains the effective g factors (gx, gy, gz) predictions for each doublet as a function of D/E Mansanares et al. (1989).Hence, the spectrum of Figure 5a shows a superposition of two main isotropic resonances and, at least, six powder spectra, for Fe 3+ ions in octahedrally and tetrahedrally coordinated sites.The powder spectra are very weak compared with the isotropic two resonance lines.For this reason, the simulation involved only the octahedrally coordinated sites.In our case, we can verify from the prediction that each line detected would be followed by other lines occurring in the region of g values between 10 and 5.For this detected line distribution in tetrahedrally distorted sites, the D/E parameter varied between 8.5 and 6.1, indicating significant rhombic distortions (obtained from the g values predictions) Mansanares et al. (1989).
Figure 6a presents the measured Q-band EPR spectrum for Cuban zeolite, showing an intense line in the center of the spectrum and another weaker, shifted to higher field.The spectrum was fitted with n = 3 in Equation 4, yielding lines with g values of 2.30 (curve 1), 2.21 (curve 2) and 2.00 (curve 3).The 2.30 and 2.21 resonances are attributed to Fe 3+ ions in octahedrally coordinate sites, probably due to ferric ions clusters from migration of extra framework Fe 3+ or even from dislodged framework ions  Goldfarb et al. (1994).The g = 2.00 line, overlapping the Fe 3+ line in the range between 12.2 kG and 12.9 kG, is the characteristic six hyperfine lines of Mn 2+ , occupying octahedral coordination Sengupta et al. (2010) as disclosed in Figure 6b.As discussed above for BRA-zeolite, here there is also no resolution to observe any rare earth magnetic ion resonance line.
As previously discussed by Franco and Rossi Franco & Rossi (2003), if the sextet do not show equally spaced lines, it points out to a random orientation of Mn 2+ ions located in environments with different crystalline fields.In our case the lines spacement differences do not exceed 3%.Another interesting aspect concerns the magnitude of the isotropic hyperfine splitting constant (A 0 ), which provides a qualitative measure of ionicity between the Mn 2+ ions and their ligands van Wieringen (1955).According to former publications Kumar et al. (1996), Chakradhar et al. (2005), this parameter can be calculated by means of the following equation: where H m is the magnetic field corresponding to m ↔ m hyperfine line, H 0 is the manganese central resonance field and m = -5/2, -3/2, -1/2, +1/2, +3/2 and +5/2.A mean value was found to be A 0 = 9.7 ± 0.7 mT, which suggests beyond an octahedral coordination due to the g ≈ 2.00, an ionic bonding character Soulayman et al. (2004), Chakradhar et al. (2005), Franco et al. (2006) of the Mn 2+ ion and its surroundings.For comparison ends, Figure 6b shows beyond the as measured spectrum (inset graph) and a simulation according to Equation 6 which includes the Zeeman term (gβH • S) and the hyperfine interaction term (A 0 I • S).
In the simulation, we adopted both the manganese nuclear spin I and the electronic spin S as 5/2.To do so, a toolbox named EasySpin Stoll & Schweiger (2006) was utilized.It permits not only a fine adjustment to the experimental data but also a baseline correction before simulation as clearly seen in Figure 6b.From the simulated curve, the parameter A 0 was found to be 9.6 mT, agreeing very well with that calculated by Equation 5 (9.7 mT).Also, the obtained g value by Equation 6, 1.99, is in close agreement with that attained by Equation 4(2.00).
From the collected data, we are also led to infer that the Mn 2+ are partially replacing the Ca 2+ ions Franco & Rossi (2003) found within the Cuban zeolite Riter (2011) without loss of symmetry.An important feature of materials used to hold nitrogen fertilizer is its feasibility of making ion exchange.NH + 4 is the typical ion present in artificial fertilizer and its incorporation inside micropores occur by its exchange with natural cations present in the zeolite structure.The presence of Mn 2+ probably contribute to the CUB zeolite to hold NH + 4 .This statement is corroborated by studies that show higher desorption rate of NH 3 from CUB zeolite than BRA one Baptista-Filho et al. (2011).

CONCLUSIONS
BRA and CUB zeolites were investigated by photoacoustic phase resolved method and electron paramagnetic resonance spectroscopy.Photoacoustic technique was useful to single out the transition bands of iron and manganese ions.The presence of both ions were confirmed through τ and τ β evaluation, whose values were found to be τ BRA = 5.40 ms, τ CUB = 4.60 ms, τ βBRA = 387 μs and τ βCUB = 305 μs.The calculated Dq-BRA /Dq-CUB = 1048 cm -1 /945 cm -1 , B-BRA /B-CUB = 457 cm -1 /813 cm -1 and C-BRA /C-CUB = 3655 cm -1 /2496 cm -1 parameters say, hitherto, that the Fe 3+ ions are both tetrahedrally and octahedrally coordinated in the Brazilian zeolite.EPR spectra revealed the characteristic g values attributed to Fe 3+ in tetrahedral symmetry accompanied by rhombic distortion (g ≈ 2.12, 2.4, 2.6, 2.8, 3.3 and 4.9 for BRA zeolite) and octahedral (g = 2.11 and 2.03 for BRA and 2.30 and 2.21 for CUB zeolites) sites.Mn 2+ ions, characterized by the six hyperfine lines, were found to be only in octahedral coordination (g = 2.00) substituting Ca 2+ ions.Furthermore, the hyperfine splitting constant with a value of A 0 = 9.7 mT points towards an ionic bonding character of the Mn 2+ ion and its vicinity.Thus, its important to emphasize that the main absorption and retention mechanism of mineral nitrogen (NH + 4 ) in zeolites is associated to the process of ion exchange, what makes the Cuban zeolite an interesting material to capture NH + 4 in substitutional Mn 2+ sites inside the micro-channels of the zeolite.

Figure 1 .
Figure 1.Vector diagram illustrating the PRM method.The resulting signal (vector R) is the sum of vectors A and B and the angles θ A and θ B stand for the phases between the signals A and B with respect to an external reference.X (in-phase signal) and Y (quadrature signal) are the projections of the vector R on the X and Y axes, respectively.
BRA-and CUB-zeolites revealed values of τ BRA = 5.40 ms, τ CUB = 4.60 ms, τ βBRA = 387 μs and τ βCUB = 305 μs, being in close agreement with reported data found in literature for relaxation times ascribed to iron Manhães et al. (2002) and manganese Baesso et al. (1989) ions.Additionally, with the help of Tanabe-Sugano diagrams Tanabe &Sugano (1954), through the diagonalization of the energy matrices for d5 configuration and adopting the same approach as

Figure 4 .
Figure 4. (a) Photoacoustic signal amplitude as a function of modulation frequency for the Cuban sample.The solid line corresponds to data fit to f −1 power law.(b) Photoacoustic phase as a function of modulation frequency for the Cuban sample.The solid line corresponds to the best fit according to Equation 2.

Figure 5 .
Figure 5. Room temperature Q-band EPR spectrum for the BRA-zeolite sample: (a) as measured spectrum and (b) low field spectrum region.