Studying the surface reaction between NiO and Al₂O₃ via total reflection EXAFS (ReflEXAFS)

The reaction between NiO and (0001)- and ()-oriented Al₂O₃ single crystals has been investigated on model experimental systems by using the ReflEXAFS technique. Depth-sensitive information is obtained by collecting data above and below the critical angle for total reflection. A systematic protocol for data analysis, based on the recently developed CARD code, was implemented, and a detailed description of the reactive systems was obtained. In particular, for ()-oriented Al₂O₃, the reaction with NiO is almost complete after heating for 6 h at 1273 K, and an almost uniform layer of spinel is found below a mixed (NiO + spinel) layer at the very upmost part of the sample. In the case of the (0001)-oriented Al₂O₃, for the same temperature and heating time, the reaction shows a lower advancement degree and a residual fraction of at least 30% NiO is detected in the ReflEXAFS spectra.

1. Introduction

Extended X-ray absorption fine-structure spectroscopy (EXAFS) is a well established method of determining the local atomic environment of a selected element in a wide variety of samples. Per se, EXAFS is a bulk technique. Surface sensitivity, however, can be achieved by recording the intensity of the reflected beam from a flat surface. In this detection mode the technique is called reflection EXAFS, or ReflEXAFS (Barchewitz et al., 1978). One of the peculiar features of ReflEXAFS is that the depth at which the sample is being probed can be controlled by varying the incidence angle of the impinging beam (d'Acapito et al., 2007). Below the critical angle (total reflection regime), the probing beam is confined as an evanescent wave in a layer just a few nanometres below the surface (Parratt, 1954). Increasing the incidence angle above the critical value, the probed depth increases sharply. However, until recently, the technique has been used mostly under the total reflection regime due to severe difficulties in data collection and analysis. Applications to buried interfaces can be retrieved in the literature, but up to now ReflEXAFS has been applied mostly to model systems. The aim of this paper is to foster the ReflEXAFS technique as a powerful tool for investigating real systems, where the structure of a thin epilayer can change with depth. In particular, attention has been focused on the chemical reactivity in the solid state at the nano-scale, having in mind, in particular, oxide systems. At the length scale of micrometres or above, the growth of a product phase is diffusion-controlled, a well established topic of the scientific literature [the basic approach (Wagner, 1936; Schmalzried, 1981, 1995) as well as more recent studies on kinetics and mechanisms of the reactivity in technologically important oxide systems such as superconductors or fuel cells components (Flor et al., 1990, 1995; Anselmi-Tamburini et al., 1991; Spinolo et al., 1992, 1993; Ghigna et al., 1993; Faaland et al., 1999; Lahl et al., 2002; Buscaglia et al., 2002; Yang & Wei, 2004; Jancar et al., 2004; Yuw & Fung, 2006; Dahl et al., 2007; Tolchard & Grande, 2007a,b; Palcut et al., 2007a,b; Waernhus et al., 2007; Kishimoto et al., 2007; Backhaus-Ricoult, 2008; Peiteado et al., 2008; Bernardo et al., 2011)]. At the nano-scale the control of a heterogeneous chemical reaction is taken by the jumps of atoms or ions across one or the other interface between product and reagents and the growth typically follows a linear time law. The most popular model oxide system investigated in this regard is the NiO + Al₂O₃ system leading to the Ni–Al spinel (NiAl₂O₄). As a matter of fact, the chemical reactivity at this level of analysis is associated with phenomena such as epitaxy and its evolution in time, strain build-up, misfit accommodation, de-wetting, volume expansion, lattice rotations, and generally with complex texture evolution and strong dependence from interfacial free enthalpy (Hesse et al., 1994; Hesse, 1997; Bench et al., 1997; Kotula & Carter, 1998a,b; Kotula et al., 1998; Johnson & Carter, 1998; Hesse & Senz, 2004; Graff et al., 2005; Farrer & Carter, 2006; Lotnyk et al., 2008; Altay et al., 2009; Arredondo et al., 2009; Basu et al., 2011). Strongly related to these aspects is the more recent use of heterogeneous reactions between nano-sized solid reagents to produce appealing nano-structures (Yin et al., 2004; Fan et al., 2006, 2007; Yang et al., 2008a,b; Qiu & Yang, 2008; Cabot et al., 2009a,b; Peng et al., 2009). Finally, going to the very early times and to the smallest length scales of a heterogeneous reaction, interest becomes focused on the processes that control nucleation, i.e. the formation of a product phase in between two grains of the reagents. This is in a sense the true chemical reaction and has been much less investigated with respect to the other reaction regimes (Kotula & Carter, 1995). Concerning the last topic, we have recently shown (Ghigna et al., 2003, 2010, 2011; d'Acapito et al., 2003a,b; Pin et al., 2009, 2011, 2013a,b) the effectiveness of an approach to the chemical reactivity in the solid state based on the use of model reacting systems (NiO + Al₂O₃ and particularly ZnO + Al₂O₃) made of a thin layer of one reagent deposited onto a single-crystal slab of the other reagent. The system is investigated by combining several experimental techniques and comparing the results obtained starting from different film-to-crystal orientations and different film thicknesses. The reaction can be quenched at different advancement degrees; the sample is allowed to react partially thus forming an epi-layer where both composition and crystal structure change with time and depth. The use of ReflEXAFS appears to be very promising in this regard, in particular due to its capability to investigate buried interfaces and multilayer systems (Heald et al., 1988), and to retrieve local structural parameters from layered systems. The reflectivity is a complex function of photon energy E as it depends on both the imaginary and real parts of the sample optical constants. Features analogous to the EXAFS oscillations are still visible above the absorption edge, but their analysis is considerably more complex. Quantitative rigorous methods for analyzing this class of data have been proposed in the past years (Heald et al., 1988; Borthen & Strehblow, 1997; Benzi et al., 2008; López-Flores et al., 2009) but have been used only in a restricted number of cases. In the past, our group has already used ReflEXAFS to study the initial steps of reactions in the solid state (d'Acapito et al., 2003a; Ghigna et al., 2003). The present contribution uses the CARD code (https://www.esrf.fr/computing/scientific/CARD/CARD.html ), which is based on the method presented by Benzi et al. (2008) and has been recently extended to the case of layered systems.

2. Experimental

2.1. Sample preparation

Thin films of NiO (25 nm thick) have been deposited onto ()- and (0001)-oriented Al₂O₃ single crystals using RF magnetron sputtering and starting from Cerac (99.995%) NiO. The deposition details have been reported elsewhere (Pin et al., 2009). An as-deposited NiO film, without any thermal treatment, was used as a reference for pure NiO (hereafter sample 1). The other deposited films have been fired at 1273 K for 6 h. For the ()-oriented single-crystal substrate (sample 2), the reaction yielding the NiAl₂O₄ spinel appears as almost complete. The chemical reaction proceeds also for the film deposited onto the other interface (sample 3), but with only a partial conversion of NiO to spinel.

2.2. ReflEXAFS data collection and analysis

ReflEXAFS data at the Ni K-edge (E = 8333 eV) have been collected at the GILDA beamline (d'Acapito et al., 1998) operative at the European Synchrotron Radiation Facility on a dedicated experimental station (d'Acapito et al., 2003b). The monochromator was equipped with a pair of Si (311) crystals and was used in dynamical focusing mode. A pair of Pd-coated mirrors (E_cutoff ≃ 18 keV) was used for harmonic rejection and for the vertical focus on the sample. The incident and reflected beams have been collected by two ion chambers, and a preliminary spectrum without the sample was used to scale the reflected intensity to absolute reflectivity values. The primary beam was shaped by a slit to 50 µm and in all the measurements it was ensured that the footprint of the beam was shorter than the sample.

For each sample, reflectivity curves at fixed energy E and variable angle φ, , were first collected at E = 8100 eV, 8400 eV and 9000 eV; these correspond to points before, just after and far after the edge, respectively.

From two angles were selected for the ReflEXAFS data collection: one before and one just after the critical angle for total reflection. The actual values were φ = 0.2° and 0.32°. This realises the desired depth-sensitive measurement, the lower-angle spectrum accounting for the surface and the higher-angle spectrum accounting for the deeper part of the sample.

As already noted, the data analysis was carried out using the CARD code. The overall procedure can be summarized as follows: as a first step, CARD fits the reflectivity to assess a model for the set of layers constituting the sample. In addition to thickness, electron density and roughness of each layer, the fitting model also accounts for instrument-related parameters, such as error in the absolute angular determination, overall normalization and background. This information is then used (a) to generate the theoretical ReflEXAFS spectra at the selected angles φ and (b) to calculate modified theoretical EXAFS paths to fit the oscillating part of each . A more thorough description of the working principle of CARD can be found in Benzi et al. (2008). It is worth noting that in the formalism used here the ReflEXAFS signal χ_R is defined as . This simple choice permits data collected either at low angle (where the edge appears with a downwards step) and at high angle (where the edge no longer possesses a step-like shape) to be treated in a coherent way. Care should be taken as the resulting χ_R spectrum appears to be reversed when compared with the standard χ_μ data (which come from spectra with an upwards step).

3. Results and discussion

Sample 1 is the pure unreacted NiO; here it is modelled as a NiO layer on an Al₂O₃ substrate. For this sample, Table 1 summarizes the fitting parameters of the reflectivity curve , whereas Fig. 1 shows the experimental data with the best fit. The thickness of the NiO layer obtained from the fit is in good agreement with the value (250 Å) determined with a quartz microbalance during the deposition process, and the density compares well with the literature value (6.67 g cm⁻³). The slight misfit at low angles is probably due to a more complex layer structure of the sample near the surface but attempts to cure it did not lead to significant improvements. On the other hand, we have verified (as also done for the other samples) that the results of the fits at the other two energy values were in good agreement with the above data.

Table 1
Fitting results of reflectivity data for sample 1 collected at 8100 eV

Data are shown in Fig. 1.

Layer	Parameter
Substrate	Roughness (Å)	9.3 (9)
NiO	Thickness (Å)	253 (25)
NiO	Roughness (Å)	38 (4)
NiO	Density (g cm^−3)	6.7 (7)

Figure 1 Reflectivity at 8100 eV for samples 1, 2 and 3. Blue line: experimental; red line: best fit. The vertical lines mark the collection angles used for the ReflEXAFS data.

The spectrum has then been modelled as shown in Fig. 2(a). From the parameters relative to the sample (thickness, density, etc.), modified EXAFS paths (Benzi et al., 2008) have been generated starting from those calculated for NiO with the Feff8.1 code (Ankudinov et al., 1998). The ReflEXAFS data χ_R(k), here defined as , have been extracted using the ATHENA code (Ravel & Newville, 2005) from the raw data, and then fitted using the ARTEMIS code (Ravel & Newville, 2005) with the modified EXAFS paths.

Figure 2 (a) for samples 1, 2 and 3 collected at 0.2°. Colours are as in Fig. 1. (b) for samples 2 and 3 collected at the highest angle. Colours are as in Fig. 1.

The ReflEXAFS spectrum is shown in Fig. 3 and at a qualitative level it closely follows the spectrum of the NiO powder. Modelling was carried out using only the first two shells (Ni–O and Ni–Ni) of the NiO local structure. The fit was carried out in r space (Fourier transforms, in Fig. 4) using a k²-weighting factor; the parameter values are shown in Table 2, and the comparison between data and fit is shown in Figs. 3 and 4. The experiment is well reproduced by the fit, and the Ni–O and Ni–Ni distances retrieved by the fit are in excellent agreement with the crystallographic data.

Table 2
Summary of the parameters for the EXAFS fitting of sample 1

Coordination numbers are the same as in the crystal structure. Errors are given in parentheses. If the error is not shown, the parameter was kept fixed to the theoretical value in the fitting.

Parameter	Theoretical	Result
E0 (eV)		−7(2)
r Ni–O (Å)	2.097	2.09 (4)
r Ni–Ni (Å)	2.966	2.97 (2)
σ^2 Ni–O (Å^2)	–	0.009 (4)
σ^2 Ni–Ni (Å^2)	–	0.007 (1)

Figure 3 ReflEXAFS spectra (blue line) with the best fit (red) collected at 0.2° and at high angle (0.32° or 0.315°) for samples 1, 2 and 3. In order to present signals with comparable amplitudes the spectra in total reflection of samples 2 and 3 were multiplied by a factor of two and those at high angles by a factor of four. The spectra of the model compounds (NiO and NiAl2O4) taken in transmission mode are multiplied by −1.

Figure 4 Fourier transforms of the ReflEXAFS spectra (blue line) with the best fit (red) collected at 0.2° and at high angle (0.32° or 0.315°) for samples 1, 2 and 3. The multiplication factors are the same as in Fig. 3.

Sample 2 is the NiO layer deposited on the () face and annealed for 6 h at 1273 K; Fig. 3 shows the experimental ReflEXAFS data along with those of the NiAl₂O₄ and NiO powder standards. For both incidence angles the spectra of this sample are strongly similar to that of NiAl₂O₄, a clear indication of the onset of the chemical reaction leading to NiAl₂O₄ and triggered by the thermal treatment. Accordingly, the layer structure of this sample is then modelled with (a) substrate, (b) a buried layer (spinel) and (c) a surface layer mixture of NiO and NiAl₂O₄. With this model the spectrum is fitted as shown in Fig. 1, while the parameters are presented in Table 3. The density of the NiAl₂O₄ layer has been kept fixed in the fit to its theoretical value (4.4 g cm⁻³), while the density of the surface layer has been left free to change. A value of 4.7 g cm⁻² has been found corresponding to a NiO and NiAl₂O₄ mixture with the ratio 10/90. data have been collected at 0.203° and at 0.320° and are shown in Fig. 2 along with their fits. ReflEXAFS data at both angles and the related best-fitting curves are shown in Figs. 3 and 4. For NiAl₂O₄, a (completely) inverse spinel structure has been used and a Ni–O shell, a Ni–Ni shell and two Ni–Al shells have been considered. The structural results from the ReflEXAFS data collected at 0.203° are shown in Table 4. Here, the advancement degree is 88%, in agreement with the reflectivity data and meaning that the reaction can be considered as complete. Note that, in order to obtain more stable fits, the σ² values have been fixed to those of the powder NiAl₂O₄ model compound taken in transmission mode. For data collected at 0.320° the fit model is a pure spinel layer; this is explained considering that, by increasing the incidence angle and the thickness of the probed layer, the surface NiO layer becomes negligible. A noteworthy result is that the parameters retrieved by the fit differ from those of bulk NiAl₂O₄, especially for what concerns the second (Ni–Ni) shell. Presumably, this indicates that at the reactive interface (the part of sample that ReflEXAFS can probe at this angle) the spinel structure is somewhat distorted, with some Ni in the tetrahedral sites or in the free octahedral sites of the spinel structure.

Table 3
Fitting results of reflectivity data for sample 2 collected at 8100 eV

Errors are given in parentheses. If the error is not shown, the parameter was kept fixed to the theoretical value in the fitting. Data are shown in Fig. 1.

Sample layer	Parameter
Substrate	Roughness (Å)	59 (6)
NiO/NiAl2O4 10/90	Thickness (Å)	64 (6)
NiO/NiAl2O4 10/90	Roughness (Å)	46 (5)
NiO/NiAl2O4 10/90	Density (g cm^−3)	4.7 (5)
NiAl2O4	Thickness (Å)	179 (18)
NiAl2O4	Roughness (Å)	59 (6)
NiAl2O4	Density (g cm^−3)	4.4

Table 4
EXAFS fit results of sample 2 at 0.203° and 0.320°

Errors are given in parentheses. If the error is not shown, the parameter was kept fixed to the theoretical value in the fitting.

	Data at 0.203°		Data at 0.320°
Parameter	Theoretical	Result	Result
NiO
r Ni–O (Å)	2.097	2.097
r Ni–Ni (Å)	2.966	3.01 (5)
NiAl2O4
r Ni–O (Å)	1.969	2.06 (2)	2.06 (2)
r Ni–Ni (Å)	2.844	2.83 (4)	3.00 (6)
r Ni–Al (Å)	2.844	2.8 (2)	2.80 (2)
r Ni–Al (Å)	3.335	3.3 (2)	3.28 (3)
For every structure
E0 (eV)		−6 (2)	3 (2)
σ^2 Ni–O (Å^2)	–	0.0073	0.016 (4)
σ^2 Ni–Ni, Ni–Al (Å^2)	–	0.0047	0.004 (1)
NiAl2O4/NiO (%)	90	88 (6)

Sample 3 is the NiO layer deposited on the (0001) face and thermally treated as sample 2. The overall layer model is here the same as for the previous sample. Fig. 1 shows the spectrum and Table 5 the parameters retrieved from the reflectivity fit whereas Fig. 2 shows the spectra. The fit of the reflectivity data now gives a surface layer containing 30% NiO and 70% NiAl₂O₄ above a pure spinel layer. When compared with the previous sample [NiO deposited onto the () single-crystal face], the fit clearly shows lower reactivity, as well as lower `roughness' and higher `thickness' of the surface layer. Moreover, as carried out for the other samples, these parameters have been used to fit the versus E spectra. The ReflEXAFS spectra (Figs. 3 and 4) exhibit a qualitative similarity to that of powder NiO at low φ, and a hybrid character at higher φ. The parameters retrieved from the fit are reported in Table 6 and confirm the qualitative data with only a 45% spinel content in the surface to be compared with the 88% found in sample 2 that confirms the lower reactivity of the (0001) face with respect to ().

Table 5
Fitting results of reflectivity data for sample 3 collected at 8100 eV

Errors are given in parentheses. If the error is not shown, the parameter was kept fixed to the theoretical value in the fitting. Data are shown in Fig. 1.

Layer	Parameter
Substrate	Roughness (Å)	29 (3)
NiO/NiAl2O4 30/70	Thickness (Å)	169 (16)
NiO/NiAl2O4 30/70	Roughness (Å)	15 (1)
NiO/NiAl2O4 30/70	Density (g cm^−3)	4.7 (5)
NiAl2O4	Thickness (Å)	82 (8)
NiAl2O4	Roughness (Å)	30 (3)
NiAl2O4	Density (g cm^−3)	4.4 (4)

Table 6
EXAFS fit results for sample 3 at 0.203° and 0.315°

Errors are given in parentheses. If the error is not shown, the parameter was kept fixed to the theoretical value in the fitting.

	Data at 0.203°		Data at 0.315°
Parameter	Theoretical	Result	Result
NiO
r Ni–O (Å)	2.097	2.097	2.097
r Ni-Ni (Å)	2.966	3.01 (5)	2.98 (3)
NiAl2O4
r Ni–O (Å)	1.969	2.06 (4)	2.05 (2)
r Ni–Ni (Å)	2.844	2.84	2.84
r Ni–Al (Å)	2.844	2.81 (4)	2.8 (1)
r Ni–Al (Å)	3.335	3.30 (5)	3.2 (1)
For every structure
E0 (eV)		−5 (2)	−1 (3)
σ^2 Ni–O (Å^2)		0.0073	0.0073
σ^2 Ni–Ni, Ni–Al (Å^2)		0.0047	0.0047
NiAl2O4/NiO (%)	70	45 (8)	61 (14)

4. Conclusion

In this work we have investigated the reaction in the solid state between NiO and Al₂O₃ using the ReflEXAFS technique on a model experimental system made of a thin NiO layer deposited onto two different orientations of single-crystal alumina substrates and allowed to react at 1273 K for 6 h. The CARD code in its new release permitted the analysis of layered samples.

It is shown that depth-sensitive information is obtained by collecting data above and below the critical angle for total reflection, thus probing (a) a surface layer a few tens of angströms thick or (b) a larger depth of the investigated sample. For the () substrate orientation, the reaction with NiO upon annealing at 1273 K is almost complete and an almost uniform layer of spinel is found below a mixed (NiO + spinel) layer at the very upmost part of the sample. A remarkable capability of the technique is that it also provides evidence of some structural distortion of the buried layer with respect to the pertinent standard phase. In the case of the (0001) substrate orientation, the reaction shows a lower advancement degree with a residual fraction of at least 30% NiO.