Probing High Oxygen Activity in YSZ Electrolyte

Reflectance spectra of Pr doped samples are given in Fig. 1a. In the SI (supplementary information), Fig. SI2, the spectra for 0.3% Pr and 5%Pr doped pellets are given. The spectra show a high reflectance background at long wavelengths with the corresponding absorption bands due to Pr³⁺ ions. These are intraconfigurational transitions from the ground state multiplet (³H₄) level to excited levels within the 4f² electronic configuration, and are clearly seen as valleys in the spectra. Broad and intense absorptions in the region from 2400 nm to 1200 nm correspond to transitions to multiplets ³H₆, ³F₂, ³F₃ and ³F₄. The band around 1000 nm corresponds to the transition to ¹G₄. The bands from 600 to 550 nm correspond to the transition to ¹D₂ and the ones with maxima from 487 nm to 450 nm to the transitions to ³P_0, ³P _1, ³P ₂. ^30,36

Zoom In Zoom Out Reset image size

At short wavelengths the reflectance diminishes as the oxygen partial pressure increases, due to absorption. The sample annealed in air shows a broad absorption around 500 nm, which can be more clearly distinguished as the dopant concentration increases (see Figs. 1b, or SI2). Therefore, it is reasonable to associate this band to absorptions involving oxidized dopant species, such as interconfigurational Pr⁴⁺ transitions or charge transfer bands (from occupied O⁼ orbitals to unoccupied Pr⁺⁴ orbitals), both of which are expected to have large optical activity (strong absorption). Hoefdraad ³⁷ assigns similar absorption bands in oxides with 8-coordinated Pr⁴⁺ to charge transfer bands, which would appear at lower energies than interconfigurational transitions.

As in the Pr doped samples, the reflectance spectra of Tb doped pellets (Figs. 2a and 2b) show a high reflectance background with strong absorption in the UV absorption edge. In the SI, Fig. SI3, the spectra for the same samples annealed in argon and 5% H₂-Ar are given, while Fig. SI4 gives the spectra for the sample YSZ-5Tb. In the near infrared (NIR) region, absorption bands (valleys) due to Tb³⁺ ions transitions within the 4f⁸ electronic configuration are clearly seen. These are absorptions due to transitions from ⁷F₆ to levels ⁷F₀, ⁷F₁ or ⁷F₂ multiplets. ^38,39 Much weaker are the Tb³⁺ absorption bands in the visible region, ⁴⁰ which are hard to distinguish in Fig. 2. Figure 3 shows the optical absorption spectra of 1at% Tb doped single crystals, in the VIS-UV range. Small absorption bands corresponding to excitations to the ⁵D₄ level at 487 nm and to the ⁵D₃ at 380 nm can be seen in the inset. It shows the absorption of the sample annealed in 5% H₂-Ar from 360 to 500 nm. Another broad absorption band in this region of the spectrum, with maximum at around 360 nm (3.45 eV) and full with at half maximum (FWHM) around 1.2 eV, is evident in the absorption spectra of the crystal annealed in air. This absorption band together with the specific scattering gives rise to the reflectance spectra as broad, diminished reflectance in the short wavelength side (λ < 500 nm), being more intense the higher the dopant concentration (see Fig. 2a) and the higher the oxygen activity of the annealing atmosphere (Fig. 2b).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Note also that the reflectance of the undoped sample decreases as wavelength increases (Fig. 2a). We can attribute this behavior to the smaller scattering as wavelength increases expected for well-sintered YSZ. This leads to a fraction of the incident light intensity escaping the experiment (either as forward transmittance or through the sample edges). As the dopant content increases, this tendency is less evident (see also Fig. SI3 in SI, for reflectance spectra of the samples annealed in 5% H₂-Ar).

The reflectance spectra for highly doped, oxidized samples show also a large absorption tail extending from the VIS up to 1200 nm, caused either by the large width of the optical absorption band or by other, less intense and broad absorptions extending towards lower energies. The optical absorption measured on a nominally 5 at% Tb doped crystal is proportional to the one shown in Fig. 3 with appropriate scaling for concentration and scattering and/or reflectance at the crystal surfaces. The spectra are shown in SI (Fig. SI5). It is clear that this absorption is due to oxidized species related to the Tb dopant, that on the same grounds than in Pr doped zirconia, can be assigned to charge transfer (from occupied O⁼ orbitals to unoccupied Tb⁺⁴ orbitals). Broad band assigned to Tb⁴⁺ CT transitions have been reported by van Vugt et al. in ZrO₂, ⁴¹ Hoefdraad in ThO₂ and other oxides, ³⁷ Verma et al. in MO-Al₂O₃ phosphors ⁴² or E. Zych et al. in Lu₂O₃ scintillators. ⁴³

Finally, note in Figs. 2b and SI4 (reflectance spectra), as well as in absorption (Fig. SI5), that the large increase in absorption in the visible–UV range associated to the formation of oxidized species is not accompanied by a decrease in absorption of the Tb³⁺ absorption bands, for example the ones in the NIR. The same is true for the Pr doped samples (see Fig. 1), and was also observed by Savoini et al. investigating the absorption spectra of single crystals. ³⁰ That is, only a very small fraction of rare earth dopant has been oxidized in the air anneals. The NIR absorption spectra of the 5% Tb doped crystal allows estimating an upper limit of 0.1% of Tb³⁺ oxidized to Tb⁴⁺ upon annealing in air at 900 °C.

The observed luminescence spectra of the YSZ doped samples annealed in air, Ar or 5%H₂-Ar flows correspond to luminescence of the dopants in the 3+ oxidation state. The luminescence spectra of Pr³⁺ and Tb³⁺ ions in YSZ have been previously reported. ^28,30,32 Here we are interested in the most intense luminescence bands in the VIS from levels with small thermal quenching. Excitation with the 488 nm Ar⁺ laser line is resonant with the ³H₄ → ³P₀ and ⁷F₆ → ⁵D₄ absorptions of Pr³⁺ or Tb³⁺ respectively, and produces luminescence bands in the visible region. In YSZ:Pr³⁺, the most intense band appears around 615 nm arising from radiative deexcitation of the ¹D₂ level to the ground state (³H₄ multiplet). In YSZ:Tb³⁺ the strongest luminescence is the one around 550 nm, produced by radiative deexcitation of the ⁵D₄ level to the first excited multiplet (⁷F₅).

The spectra of samples with 1 at% and 5 at% doping of the respective ions, and annealed in the different atmospheres at 900 °C are given in Fig. 4 (YSZ-xPr) and Fig. 5 (YSZ-xTb). Quantitative details are collected in Table II. For both dopants, the samples annealed in reducing atmosphere show the strongest luminescence. The intensity ratio of each sample vs its value in the reduced sample is also given in Table II. This ratio changes consistently (above the uncertainty of the measurement) and is smaller in the samples with 5 at% Tb. Note for example that the measured luminescence in the samples doped with 5 at% Tb annealed in air are around 1/3 of the one measured for the samples annealed in H₂. For 1 at% Tb or Pr doped samples the ratio is around 0.7 to 0.8.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The dependence of the average peak intensity with concentration is different for each dopant (Table III). On the one side, YSZ-5Pr samples show smaller luminescence intensity than YSZ-1Pr, as well as a change in the shape of the spectra. Savoini et al. ³⁰ observed a faster decay monitoring the ¹D₂ to ³H₄ luminescence in YSZ doped with 6.3 at% Pr, ascribed to concentration quenching by cross-relaxation, ⁴⁴ which explains the strong decrease of the luminescence intensity, as also recently observed by Wang et al. ⁴⁵ for YSZ doped with Pr at concentrations above 1 at% Pr. This luminescence band also overlaps the rather intense low energy feature of the absorption corresponding to the transition ³H₄ to ¹D₂. ⁴⁶ We did not use confocal diaphragm to register the luminescence, so that light emitted at planes below the surface comes inside the spectrograph and CCD detector. The attenuation of the emitted light by self-absorption can contribute to the change of the luminescence band profile of sample YSZ-5Pr with respect to the one of YSZ-1Pr. In addition, other luminescence lines from deexcitation of the ³P₀ to the ³H₆ multiplets might become apparent as the intensity of the ¹D₂ to ³H₄ luminescence decreases by cross-relaxation, and thus change the shape of the spectrum. Note that this luminescence intensity ratio is only slightly dependent on the thermochemical treatment, compatible with processes related to Pr³⁺ to be the responsible ones for the change in intensity.

In the Tb doped samples there is a very slight broadening of the spectra and a shifting of the maximum towards higher energies (by 4 cm⁻¹ approx.) with the increase in the Tb concentration. There is also a more intense luminescence as the Tb concentration increases. This is clearly seen in Fig. 6, where the intensity values at maximum have been plotted with respect to the Tb dopant concentration. This luminescence is not expected to be strongly affected by concentration quenching. ^47,48 The dependence is sublinear, that is, the measured luminescence intensity increases slower than the Tb concentration.

Zoom In Zoom Out Reset image size

NIR reflectance allows a rough upper limit estimate of the proportion of Tb ions oxidized from the NIR reflectance spectra, giving that less than 1% of Tb ions were oxidized by the annealing in 100 bar O₂. Therefore, a relative quantification of the Tb⁴⁺ content will be possible following the variation of the optical signals.

In the absence of light emission, the measured total hemispherical reflectance (Fig. 7) is the amount of the incident light intensity that is neither absorbed nor transmitted by the sample, that is,

$$\begin{eqnarray}&&1={\rm{R}}+{\rm{T}}+{\rm{A}}\end{eqnarray} \tag{ 1 }$$

where R, T and A being total hemispherical reflectance, forward transmittance and absorptance. Outside the absorption bands, the forward transmittance increases with wavelength due to decreasing scattering. One can take out transmitted light from the equation (and the attenuation of the rays that end up forward transmitted) by dividing the measured reflectance by the reflectance that would be measured in a sample without absorption, that we will call R_white. We can estimate R_white with the dotted straight line plotted in Fig. 7a. R_norm is calculated as

$$\begin{eqnarray}&&{{\rm{R}}}_{{\rm{norm}}}={\rm{R}}/{{\rm{R}}}_{{\rm{white}}}\end{eqnarray} \tag{ 2 }$$

and plotted in Fig. SI5. Now, we approximate the above Eq. 1 to

$$\begin{eqnarray}&&1={{\rm{R}}}_{{\rm{norm}}}+{\rm{A}}\end{eqnarray} \tag{ 3 }$$

where R_norm is the normalized reflectance as described, and A is the absorptance.

For small enough absorption coefficient it is reasonable to assume that the volume of sample explored in the reflectance experiment at each wavelength is determined by the microstructure (and possibly sample thickness according to the present normalization) and is independent of the absorption. Upon this assumption, from A it is possible to estimate the optical density (OD) or absorbance with

$$\begin{eqnarray}&&{\rm{A}}=1-{10}^{-{\rm{OD}}}\end{eqnarray} \tag{ 4a }$$

or

$$\begin{eqnarray}&&{{\rm{R}}}_{{\rm{norm}}}={10}^{-{\rm{OD}}}\end{eqnarray} \tag{ 4b }$$

OD is proportional to the absorption coefficient and the length of the light path in the material. If the length of the light path is constant, OD is a measurement of the absorption coefficient, which is the product of the molar absorption coefficient times the concentration of the optically active species. We have the measurement of the Tb⁴⁺ relative concentration.

The values of the calculated OD for selected wavelengths as a function of P_O2 are given in Fig. 8. On the same plot, with open symbols, we give also the OD estimated for a 2 mm thick YSZ crystal doped with 3 at% Tb. These have been estimated from the OD measured for YSZ-5Tb crystals annealed at 900 °C in Ar or air (see Figs. 3 and SI5) as proportional to the respective dopant content. Note that at λ = 500 nm, both estimates coincide nicely for samples annealed in Ar (absorption coefficient ≈ 0.52 cm⁻¹) or air (absorption coefficient ≈ 3.0 cm⁻¹). This means that 2 mm is a reasonable estimate for the 500 nm light path length in the ceramics in the course of the reflectance measurement. The reasonable agreement between both estimates of absorption at 500 nm but large disagreement at 400 nm, tells also that in the latter case, the estimate of the OD from reflectance measurements as proportional to the absorption coefficient is valid up to OD values of around 0.15 (absorption coefficient of the crystal smaller than 2 cm⁻¹).

Zoom In Zoom Out Reset image size

At 750 nm and 600 nm, the assumption of small absorption coefficient (α) holds. From the absorption spectra in a sample annealed in air (Figs. 3 or SI5) one can infer that it is smaller than α = 0.3 cm⁻¹, so that, after a light path (d) of around 2 mm, OD = 0.434αd would be below 0.03. All OD values in Fig. 8 for 600 and 750 nm are below 0.1. As indicated by the lines, this OD is approximately proportional to ${P}_{{O}_{2}}^{1/4}.$ At shorter wavelengths (larger optical absorption), the proportionality of OD with ${P}_{{O}_{2}}^{1/4}$ is observed for the measurements in the single crystals (λ = 400 nm or 500 nm), or for the estimates in the ceramic samples when OD remains below around 0.15.

At large OD, for example for λ = 500 nm and very oxidized samples or for λ = 400 nm, there is a deviation of the ${P}_{{O}_{2}}^{1/4},$ that might be expected as the absorption coefficient cannot be so easily extracted from R_norm, or otherwise stated, the OD value estimated as –log (R_norm) is no longer proportional to the absorption coefficient. The light path decreases as the optical absorption coefficient increases.

The proportionality of the absorption coefficient with ${P}_{{O}_{2}}^{1/4}$ is in agreement with the absorption being due to oxidized species trapping one electron hole. In YSZ, an oxide ion conductor with large concentration of oxygen vacancies, the incorporation of oxygen fills-in a small proportion of oxygen vacancies and introduces electron holes. The latter ones oxidize Tb³⁺ ions to Tb⁴⁺, ultimately responsible for the observed absorption. The concentration of Tb⁴⁺ species depends on the P_O2 . ⁴⁹ The equilibrium with the atmosphere through oxygen exchange

$$\begin{eqnarray}&&\displaystyle \frac{1}{2}{O}_{2}+{V}_{O}^{\cdot \cdot }\rightleftarrows {O}_{O}^{\times }+2{h}^{\cdot }\end{eqnarray} \tag{ 5a }$$

for constant concentration of oxygen vacancies and oxide ions states that

$$\begin{eqnarray}&&\left[{h}^{\cdot }\right]\propto {\left({p}_{{O}_{2}}\right)}^{1/4}\end{eqnarray} \tag{ 5b }$$

Local redox equilibrium (hole trapping)

$$\begin{eqnarray}&&T{b}_{Zr}^{\times }\rightleftarrows {{\rm{Tb}}}_{Zr}^{/}+{h}^{\cdot }\end{eqnarray} \tag{ 6a }$$

tells that

$$\begin{eqnarray}&&\left[T{b}_{Zr}^{\times }\right]\propto \left[T{b}_{Zr}^{/}\right]\left[{h}^{\cdot }\right].\end{eqnarray} \tag{ 6b }$$

From Eqs. 5b and 6b it results,

$$\begin{eqnarray}&&\displaystyle \frac{\left[T{b}_{Zr}^{X}\right]}{{\left[T{b}_{Zr}\right]}_{0}}=\displaystyle \frac{{P}_{{O}_{2}}^{1/4}}{{P}_{0}^{1/4}+{P}_{{O}_{2}}^{1/4}}\end{eqnarray} \tag{ 7 }$$

Where ${\left[T{b}_{Zr}\right]}_{0}$ is the total concentration of Tb, and P₀ is a temperature dependent parameter that encloses the proportionality constants of Eqs. 5b and 6b. We have used Kröger-Vink notation.

In the present case where only a negligible proportion of Tb is oxidized, $\tfrac{\left[T{b}_{Zr}^{\times }\right]}{{\left[T{b}_{Zr}\right]}_{0}}$ ≪1,

$$\begin{eqnarray}&&\displaystyle \frac{\left[T{b}_{Zr}^{\times }\right]}{{\left[T{b}_{Zr}\right]}_{0}}={\left(\displaystyle \frac{{P}_{{O}_{2}}}{{P}_{0}}\right)}^{1/4}\end{eqnarray} \tag{ 8 }$$

P₀ , the pressure at which the concentration of reduced and oxidized Tb species is equal, takes in this case extremely high values, higher than 10⁸ bar.

The luminescence intensity values at the peak maximum (543.6 nm) as a function of oxygen partial pressure of thermochemical treatment are given in Fig. 9. As already argued above, the measured Tb³⁺ luminescence is excited with the amount of incident light intensity that is available after reflection at the incident sample surface and absorption by the oxidized species into the volume of sample intervening in the experiment. Therefore, only the non-absorbed light is available for excitation, or 1−A = R_norm if one would measure the back scattered reflectance in the optical configuration used to measure the luminescence. As above, R_norm can be written as 10^−OD. Let us assume that the participating volume is small enough such that it is the same at all conditions (small optical density also in the most oxidized sample). This assumption is expected reasonable for excitation and collection through a microscope objective, and it results in OD (λ₁) being proportional to the absorption coefficient and therefore to the concentration of the oxidized species, $\left[T{b}_{Zr}^{X}\right].$ λ₁ is the excitation wavelength, 488 nm. Similarly, the emitted light, with wavelength λ₂ = 543.6 nm, suffers scattering and absorption before reaching the objective towards the detector. After reflection, scattering and absorption, the luminescence intensity will be proportional to

$$\begin{eqnarray}&&{I}_{lumin}\propto {I}_{0}\cdot {10}^{-OD\left({\lambda }_{1}\right)}\cdot {10}^{-OD({\lambda }_{2})}\end{eqnarray} \tag{ 9 }$$

where we have assumed that the proportion of light reflected at the incident surface is constant, independent of the sample oxidation state. Both optical densities are proportional to ${P}_{{O}_{2}}^{1/4},$ and therefore a functional dependence such as

$$\begin{eqnarray}&&{I}_{lumin}\propto {10}^{-b\cdot {P}_{{O}_{2}}^{1/4}}\end{eqnarray} \tag{ 10 }$$

is expected. To better fit to the experimental data, a constant independent term has been added. The continuous line in Fig. 9 is a fit to the data with the expression

$$\begin{eqnarray}&&\displaystyle \frac{{I}_{lumin}}{{I}_{{H}_{2}}}=a+(1-a)\cdot {10}^{-b\cdot {P}_{{O}_{2}}^{1/4}}\end{eqnarray} \tag{ 11 }$$

which gives a = 0.124 ± 0.014 and b = 0.41 ± 0.03 bar^−1/4, with R = 0.9964 and P_O2 given in bar. Figures after ± are standard errors. This rough approximation cannot explain the extra term needed to fit the data. b has to be compared with the known values of extinction coefficients. From the reflectance spectra (Figs. 7 or SI5) of the sample annealed in Ar (low absorption), one can estimate OD (λ₁)/OD (λ₂) = 2.5. Then one can split the total OD (λ₁) + OD (λ₂) = $b\cdot {P}_{{O}_{2}}^{1/4}$ between both wavelengths. Let us take ${P}_{{O}_{2}}$ as 0.21 bar. Then OD (λ₁) + OD (λ₂) = 0.28, and OD (λ₁) = 0.20. From the single crystal optical absorption spectra we can estimate the absorption coefficient as 3.4 cm⁻¹ at 488 nm for 3Tb-YSZ. With this value, one can estimate the path-length the light has travelled in the sample in the luminescence experiment as

$$\begin{eqnarray*}&&d=\displaystyle \frac{OD\left({\lambda }_{1}\right)}{0.434\alpha \left({\lambda }_{1}\right)}=1.4{\rm{mm}}\end{eqnarray*}$$

If we force to a = 1 in the fitting, b decreases to 0.31 and the estimate for d goes to 1.1 mm. This estimate for d is near to the one estimated in the reflectance experiments, although somewhat smaller as expected by the different optical arrangement. The actual distance that the light has travelled into the sample would be much smaller than 1.1 mm due to scattering. For example, in dense Al₂O₃-YSZ-YAG micro-composites produced by controlled solidification of eutectics, Mesa et al. ⁵⁰ estimated 5 times longer path-length than sample thickness. A better knowledge of the light scattering in these ceramic pellets, which contain tiny residual porosity, is needed to know the real volume that participates in the experiment. Based on the present results, the depth of sample explored by the backscattering luminescence experiments might be of the order of 100 μm.

Zoom In Zoom Out Reset image size

It is relevant to note that in the range from 10⁻⁴ to 100 bar, there is an almost linear relationship between luminescence intensity and ${P}_{{O}_{2}}^{1/4}$ that can be used to monitor the oxygen activity in these YSZ electrolytes.