Accurate and Precise Measurement of Oxygen Isotopic Fractions and

Feb 3, 2015 - International Institute for Carbon Neutral Energy Research ... In this work, the selective attenuation of secondary ion signals (SASI) a...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/ac

Accurate and Precise Measurement of Oxygen Isotopic Fractions and Diffusion Profiles by Selective Attenuation of Secondary Ions (SASI) Helena Téllez,*,†,⊥ John Druce,†,⊥ Jong-Eun Hong,‡,∥ Tatsumi Ishihara,†,‡ and John A. Kilner†,§ †

International Institute for Carbon Neutral Energy Research (WPI-I2CNER), Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan ‡ Department of Applied Chemistry, Kyushu University, 744 Motooka, Nishi-ku, Fukuoka 819-0395, Japan § Department of Materials, Imperial College London, Prince Consort Road, South Kensington, London SW7 2BP, United Kingdom ABSTRACT: The accuracy and precision of isotopic analysis in Time-of-Flight secondary ion mass spectrometry (ToF-SIMS) relies on the appropriate reduction of the dead-time and detector saturation effects, especially when analyzing species with high ion yields or present in high concentrations. Conventional approaches to avoid these problems are based on Poisson dead-time correction and/or an overall decrease of the total secondary ion intensity by reducing the target current. This ultimately leads to poor detection limits for the minor isotopes and high uncertainties of the measured isotopic ratios. An alternative strategy consists of the attenuation of those specific secondary ions that saturate the detector, providing an effective extension of the linear dynamic range. In this work, the selective attenuation of secondary ion signals (SASI) approach is applied to the study of oxygen transport properties in electroceramic materials by isotopic labeling with stable 18O tracer and ToF-SIMS depth profiling. The better analytical performance in terms of accuracy and precision allowed a more reliable determination of the oxygen surface exchange and diffusion coefficients while maintaining good mass resolution and limits of detection for other minor secondary ion species. This improvement is especially relevant to understand the ionic transport mechanisms and properties of solid materials, such as the parallel diffusion pathways (e.g., oxygen diffusion through bulk, grain boundary, or dislocations) in electroceramic materials with relevant applications in energy storage and conversion devices.

T

as secondary ion mass spectrometry (SIMS),3,4 nuclear reaction analysis (NRA),5−7 or Raman spectroscopy.8,9 The former (“gas phase”) approaches have attracted recent interest,10,11 and enjoy a particular strength that by monitoring the evolution of the different oxygen isotopes (i.e., 16O16O, 16 18 O O, 18O18O), mechanistic information can be inferred, particularly on the relative rates of dissociation versus incorporation, assuming the individual steps in the oxygen exchange reaction can be lumped into a two step reaction. However, this type of technique shows poor sensitivity to D, mainly because the size of the particles in the powders. Samples in powder form are preferred as they maximize the surface area, and hence, sensitivity to k is well below the “critical length”, Lc = D/k, meaning that the measurements are typically in a pure surface exchange-limited regime, rather than mixed diffusion and surface exchange limited transport kinetics. Conversely, solid-state methods are performed under a mixed kinetic limitation condition, representing a sensitive methodology for the determination of both D* and k* values, as long as appropriate boundary conditions are maintained during the oxygen exchange experiment.12−15 Moreover, these

he measurement of the kinetics of oxygen surface exchange and diffusion in oxide materials is critical for characterizing the transport properties of advanced electroceramic materials with significant applications as sensors, oxygen separation membranes, solid oxide fuel cells and electrolyzers (SOFC and SOEC, respectively). There are three general classes of measurements of the oxygen diffusion (D) and surface exchange (k) coefficients, which characterize the oxygen transport properties of such materials.1 These are (i) chemical diffusion or relaxation methods, (ii) direct electrochemical methods, and (iii) stable isotope exchange methods. Stable isotope techniques ideally measure a selfexchange of oxygen which is driven purely entropically (i.e., in the absence of any gradients in chemical potential). The derived parameters, termed tracer diffusion and surface exchange coefficients, and commonly denoted with an asterisk, (D* and k*, respectively), are comparable to self-diffusion or -exchange coefficients, within a correlation factor of the order of unity (e.g., 0.69 for ideal cubic perovskite materials).2 Measurements of oxygen exchange by isotopic techniques can be further broken down into “gas phase” techniques, in which the evolution of the 18O isotopic fraction in the gas is recorded as a function of time, and “solid-state” techniques, in which the isotope is diffused into the solid and the quenched lateral distribution recorded by an isotopically sensitive technique such © XXXX American Chemical Society

Received: November 26, 2014 Accepted: February 3, 2015

A

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

heating at 80 °C until it formed a viscous gel. The as-prepared gel was heated at 180 °C, which led to self-ignited combustion. After heating the ash at 400 °C for 2 h, it was pulverized and calcined at 1000 °C for 6 h. The prepared powder was pressed into pellets and then sintered at 1500 °C for 10 h. The phase purity of the polycrystalline powders was checked by XRD diffraction confirming the rhombohedral perovskite structure. The LSM20 pellets were ground flat using SiC grinding paper, and a smooth surface was obtained by polishing with a waterbased diamond suspension to a final particle size of 1/4 μm (final surface roughness of around 60 nm). Any residue from the polishing was removed by sequentially cleaning the sample in an ultrasonic bath with acetone, ethanol, and deionized water for 10 min in each solvent. The relative density of the pellets as measured by the Archimedes method was 95%. Isotope Exchange. Isotope exchange experiments were performed using a custom-built exchange setup following the procedure described elsewhere.3,4 The exchange tube was evacuated to 30 nA maximized for best signal statistics 0.6−5 μm

mass resolution

∼11 000

burst alignment or imaging mode (BA) focused by the lens source at the 2nd aperture 30 keV 0.4−0.7 nA reduced by beam bursting Bi+: 150 nm Bi3+: 120 nm 1 (non burst) > 6000 (burst)

collimated burst alignment mode (CBA)33 collimated at the 1st aperture and focused at the 2nd aperture 25 keV 70−100 pA adjusted to avoid signal saturation Bi3+2< 100 nm

Figure 1. Dependence of (a) raw and (b) corrected (Poisson and attenuation factor) 16O secondary ion intensity as a function of primary ion pulse width and corresponding time-averaged primary current measured with 95 μs cycle time.

1 (no burst) > 6000 (burst)

otherwise. The burst alignment or fast imaging mode (BA) was performed using eight pulses following the experimental details previously described.31 Finally, a further decrease of the beam current was achieved by collimating the beam through aperture 1 and focusing the beam further using the second of the three lenses in the ion column (lens magnification), as described by Holzlechner et al. (CBA mode).32 A detailed description of the primary ion source modes can be found elsewhere.32,33,36

Therefore, SASI allows the determination of reliable isotopic fractions by attenuating the intensity for the high yield SI reaching the detector while using large pulse widths to minimize the effect of the beam instability and reduce statistical uncertainties associated with low SI yield species when using low PI currents. In this way, the region in which the 16O signal is linear is extended toward larger pulse widths when SASI is applied (i.e., nonlinearity of 16O signals starts from 7 and 11 ns for low and high attenuation levels, respectively). Moreover, full saturation of the detector is observed for pulse widths larger than 10 ns for low SASI (Figure 1a, red circles). In contrast, the high SASI series does not show full saturation in the range of primary ion currents measured, although there is a nonlinear dependency for high pulse widths (Figure 1a, blue triangles). Some of this nonlinearity may be corrected by accounting for Poissonian counting statistics.30 We note that it is important that any Poisson correction is applied before scaling by the SASI factor; the counting experiment here is upon the attenuated signal. The resulting corrected signals, which have also been scaled by the SASI factor, are shown in Figure1b. From this figure, it is clear that the application of Poisson correction to minimize dead-time effects extends the linear response regions, although the nonattenuated and low-SASI series still fully saturate at relatively low primary pulse widths (16O signal saturation levels are indicated by dashed lines for each series). The Poisson correction procedure ameliorates the nonlinearity of the series recorded with lower attenuation, which remains linear until 0.78 pA (11 ns pulse width), although it is unable to correct the counts once the detector is fully saturated (primary current greater than 0.92 pA @ 12 ns pulse width). In contrast to these results, the Poisson-corrected high SASI series showed a linear response along the whole range of PI currents, reflecting the true intensity of the 16O secondary ions. The appropriate SASI level is, however, sample-specific and must be selected taking into account the isotopic fraction of the species of interest, its expected concentration range through the analysis, and other analytical requirements, such as accuracy and precision. Figure 2 illustrates the effects of different SASI levels applied in HCBM mode for the analysis of 16O and 18O isotopes in a LSM20 ceramic sample with natural isotopic abundance. The high PI current used in HCBM assures maximum beam stability while the SASI levels were changed during the analysis for each oxygen isotope. Both raw and corrected profiles are plotted for the oxygen isotopes. The



RESULTS AND DISCUSSION Depending on the specific analysis requirements for a particular measurement, the primary ion (PI) source can be operated in several modes, as described in Table 1. For those applications in which the lateral resolution is not the main requirement, but high mass resolution is desirable, the PI source is typically run in HCBM mode for the spectrometric analysis of the surface composition. In HCBM, a package of primary ions produced at the ion source is mass-selected, and its current is adjusted by a beam chopper located after the lens magnification. This current adjustment is performed by selecting the appropriate “pulse width” values (i.e., the current of the PI pulse is effectively increased with the primary beam pulse width). However, it should be noted that the pulse duration at the target surface is virtually independent of the pulse width for the standard range used in HCBM (i.e., less than 50 ns) because the PI pulse is later compressed by a pulse buncher at the end of the PI optical train. In the case of species with high SI yields or present in high concentrations, as in the case of 16O in oxides, the use of the HCBM eventually leads to detector saturation, even for very short pulse widths. Figure 1a shows the raw intensity of the 16 − O SI peak in a LSM ceramic sample as a function of PI pulse width in HCBM when no signal attenuation is applied (Figure 1a, black squares). The 16O SI intensity increases sharply with PI current until the detector saturates (saturation point indicated as a black dashed line in Figure 1a). A straightforward way to avoid detector saturation is to reduce the pulse width and, hence, the PI current applied. However, as seen in Figure 1, this would involve a significant decrease of the pulse widths, making the detection of the minor oxygen isotopes very difficult (note that the 16O signal saturates even for low PI pulse widths, i.e., 5 ns pulse width corresponding to a PI current of 0.055 pA). By applying appropriate SASI, the effective dynamic range is extended to allow the detection of both high and low yield species. C

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Figure 3. Average 18O isotopic fraction values obtained in LSM20 annealed in air for different analysis modes. HCBM/spectrometry mode (blue squares; A2−A7: see different SASI levels applied in Figure 2), BA/fast imaging mode (red spheres; B1:8 bursts, B2: ignoring burst mode for 16O) and CBA (black triangles; C1: no bursting, C2: burst beam). The reference value for δ(18O) as reported by NIST is indicated as a gray region.

Figure 2. ToF-SIMS signal intensities obtained for 16O and 18O secondary ions analyzed in HCBM with no attenuation (open symbols) and after Poisson correction and SASI (closed symbols) in a LSM20 sample sintered in air. The 18O isotopic fraction is also shown as a solid blue line.

corrected profiles are obtained by first applying Poisson correction to the raw intensities, followed by scaling by the appropriate attenuation factor. The segments in which the 16O signal was unattenuated (A1, A8, and A9), showed a flat measurement due to detector saturation for this mass channel (open black spheres), which cannot be fully corrected even using Poisson correction (closed black spheres). The detector saturation in these segments led to systematic errors in the 18O isotopic fraction, with estimated values of 1.6% almost 1 order of magnitude higher than the natural 18O abundance of 0.205% reported by NIST.37 On the other hand, no detector saturation occurs for the nonattenuated signal at m/z 18, given the low concentration of the 18O isotope (segments A1−A3). Therefore, the raw (open red triangles) and the corrected signals (closed red triangles) show the same intensities for the 18O signal in these sections. In those segments where SASI was applied for 18O (A4-A9), the signal statistics became rather poor due to the weak intensity and hence the low signal-to-noise ratio. This is especially so when using the high SASI, which results in a very low absolute level of counts (i.e., only around 20 counts distributed across a single 128 × 128 pixel frame of the measurement). A better comparison of the effect of the applied attenuation on the accuracy and precision of the apparent 18O isotopic fraction is shown in Figure 3, which presents the average of the isotopic fraction values in each region, along with their standard deviations. The 18O natural abundance as provided by NIST is plotted for comparison (δ(18O) = 0.205 ± 0.014%, gray region).37 Those segments with no attenuation of the 16O signal and subject to systematic errors due to detector saturation are not shown (i.e., A1, A8, A9). We also see from Figure 3 that the low SASI for 16O is not enough to avoid saturation of this signal in this measurement, leading to an overestimation of C′x for the respective regions (A2, A4, A7). Moreover, the low (i.e., background) level of 18O counts leads to larger standard deviations in the apparent isotopic fractions when SASI is applied (A4−A7 segments). Hence, signal attenuation is necessary only for the 16O species in order to avoid saturation and ensure accurate determination of the isotopic fraction; application of signal attenuation to the 18O SI only degrades the precision of the measurement due to poorer counting statistics. We note that the most accurate and precise 18 O isotopic fraction is determined when high SASI is applied

to the 16O and the 18O is not attenuated (A3), as the segment with low attenuation for 16O (A2) still shows a slightly increased apparent 18O fraction due to slight saturation of the detector (as observed for the corrected 16O signal in comparison with A3 in Figure 2). Figure 3 also shows the apparent 18O isotopic fractions determined using other primary ion source alignment modes typically applied to reduce the PI currents and avoid dead-time effects. The values for B1 and B2 were obtained using the conventional “burst mode”, as described by De Souza et al.31 For B1, all eight bursts were used to determine C′x; this leads to a higher apparent value due to some saturation of the detector by the arrival of the first burst of 16O ions (this is evident in the mass spectrum as the subsequent seven pulses were lower in intensity than the first one). As described in ref 31, a simple way to allow for this deviation is to use the Poisson-corrected intensity from only the first pulse and multiply by the total number of pulses. In this way, a more accurate determination of the 18O isotopic fraction was obtained (B2). Another methodology is the Collimated Burst Alignment mode (CBA) that provides low PI current but unit mass resolution.32 This can be improved by bursting the beam in a similar way to that used in the “burst mode” measurements. This also has the effect of further decreasing the time-averaged primary current. However, both nonburst (C1) and burst (C2) CBA modes led to less accurate 18O isotopic fractions in the LSM sample compared to the conventional BA and the SASI modes. The nonburst (C1) CBA measurement appears still to show some detector saturation, giving a systematically high estimation of the 18O isotopic fraction. However, bursting reduces the intensity of both signals and, hence, decreasing the signal-to-noise ratio and increasing the uncertainty of the 18O isotopic fraction. It would be possible to mitigate the saturation of 16O which leads to the inaccurate value for C1 by further reducing the PI current (note we used a pulsed current of 0.068 pA@50 μs, which is at the low end of the range 0.03−0.05 pA @100 μs specified by Holzlechner et al.).32 However, decreasing the PI current further would decrease the precision of the measurement of 18O SI intensity, adversely affecting the precision of the measured C1. D

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry Both BA and CBA modes avoid dead-time effects in the detector by decreasing the PI intensity, which enables the detection of the species with high ion yields and/or high concentration, such as 16O, by globally decreasing the intensity of the spectrum. It must be noted that these methodologies do not extend the dynamic range of the analysis as all the SI signal will be decreased in proportion to the decrease in PI intensity. This often leads to larger dispersions of the measured 18O isotopic fractions due to the poor counting statistics resulting from weak 18O SI intensities. Conversely, the use of the SASI attenuation in combination with the HCBM ion source alignment avoids saturation while maintaining the intensity of the weaker peaks; hence it provides a real extension of the dynamic range, maintaining good limits of detection for low yield species. Further advantages of the SASI mode over the reduced current modes (BA and CBA) are straightforward operation and alignment procedures, high mass resolving power to avoid systematic errors due to isobaric interferences (e.g., 16 − 32 2− O , S , CH4−), and the reduction of the acquisition time for imaging analysis due to the improved signal statistics. In the next set of experiments, we applied the SASI methodology to the analysis of oxygen tracer diffusion in La0.8Sr0.2MnO3 (LSM20). This electronically conducting perovskite is widely used as a cathode for solid-oxide fuel cells.38,39 The oxygen transport properties of LSM20 have been studied using the so-called isotope-exchange depth profiling technique,34 showing a surface exchange coefficient at 1000 °C (k*) 2 orders of magnitude higher than the bulk diffusion coefficient (D*). For practical annealing times, this implies high values of the parameter h′ =

k∗ D∗t D∗

Figure 4. ToF-SIMS depth profiles for 16O, 18O signals (raw and corrected) as obtained using (a) automatic SASI in HCBM mode and (b) BA mode with eight bursts. Normalized 18O isotopic fraction profiles (Cx′) estimated from the ToF-SIMS profiles: (c) automatic SASI in HCBM mode and (b) BA mode with eight bursts. The 18O bulk diffusion contribution as obtained from the fitting is also shown (blue dotted lines).

triangles), three different regions are observed: a high attenuation is necessary until 0.9 μm, where the SASI is switched to the lower level. Finally, no attenuation is applied to the 18O SI for depths larger than 2.75 μm, as the isotopic fraction approaches the natural abundance. Note that at this point, high SASI is necessary to avoid saturation of the 16O signal. However, because this attenuation is only a local decrease of the selected peak and not a global reduction of the total SI intensities, we are able to simultaneously detect the 18O at nearly background abundance with good counting statistics. The corrected intensities for both oxygen isotopes in BA mode are significantly lower than in the SASI-HCBM mode, and as a result, the signal for 18O is noisier at the end of the BA mode profile (Figure 4b, closed red triangles). The 18O isotopic fraction at a distance x from the surface (Cx) is given by

(1) 18

which in turn means the O isotopic fraction at the surface equilibrates very quickly with the gas phase.3 LSM20 represents, then, an ideal material to show the applicability of the new methodology for the determination of 18O diffusion profiles, because the isotopic fraction decreases from almost 1 at the surface (18O isotopic fraction in the gas phase is 0.95 in our experiment) to close to the natural abundance over a short distance. Therefore, the SASI factor applied must be also tuned along the depth profile in order to avoid both dead time effects for the 18O at the start of the profile and the 16O deeper in the sample, but also poor signal statistics due to the weak 18O intensity at the end of the diffusion profile. Figure 4 shows the 16O and 18O oxygen profiles measured using SASI modulated automatically by the instrument software in HCBM (Figure 4a) and for the “conventional” BA mode (Figure 4b). Although the BA mode uses a reduced PI current the 16O signal still saturates, as observed in the mass spectrum (the first burst of the 16O peaks showed higher intensity compared to the following seven bursts). To avoid such saturation, we used the first 16O burst intensity and scale to the number of bursts applied (e.g., eight bursts).31 On the other hand, the different SASI levels applied along the depth profile are easily observed in the raw signals in Figure 4a (open symbols). The 16O raw profile (open black circles) starts with a low SASI (the difference between the raw and corrected profiles is around 1 order of magnitude), but it switches very rapidly to high SASI at around 0.1 μm. This is seen as a sharp drop in the raw intensity (open circles), whereas the corrected profile, accounting for the different SASI levels, remains at the same level. In the case of the 18O secondary ions (open red

Cx =

I(18 O) I(18 O) + I(16 O) 18

(2)

16

where I( O) and I( O) are the corrected SI intensities after applying Poisson correction and scaling by the corresponding SASI level. This apparent 18O isotopic fraction must be normalized to 18O concentration at the gas phase (Cg = 95% in our experiment) and at background (Cbg = 0.21%, corresponding to the 18O isotopic fraction of the gas during the equilibration step). The normalized 18O isotopic fraction (C′x) can be then fitted using Crank’s solution to the diffusion equation for a semi-infinite medium under surface evaporation conditions.40 In this case, the oxygen penetration profile in LSM is expected to show a significant component attributed to grain boundary (gb) diffusion, as described previously by De Souza et al.34 Therefore, an additional tailing factor has been included to account for this contribution (with the corresponding grain boundary tailing function parameters Agb and Zgb), as shown in eq 3: E

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry Cx′ =

Cx − C bg Cg − C bg

⎡ ⎤ ⎡ ⎛ * 2 ⎞ x ⎥ − ⎢exp⎜ k ·x + (k*) ·t ⎟ · = erfc⎢ * ⎢ 2 D *·t ⎥ ⎢ ⎝ D b* Db ⎠ ⎣ ⎣ b ⎦

⎛ ⎞⎤ x t ⎟⎥ erfc⎜⎜ + k∗· + [Agb ·exp(−Zgb·x 6/5)] D b* ⎟⎠⎥⎦ ⎝ 2 D b*·t

rearrangements that could take place during annealing at high temperatures.41,42 This will reduce significantly the analysis time to obtain a complete picture of all the diffusion and segregation processes taking place during the annealing and exchange. Figure 5 shows the 18O normalized isotopic fractions calculated from the M16O− and M18O− profiles. The calculated C′x isotopic fractions from the MO− profiles are very similar to those obtained from the oxygen isotopes (i.e., 16O− and 18O−) for both analysis modes (SASI and BA modes). However, the larger dispersion in C′x obtained for the BA analyses (Figure 5b) is due to weaker MO− signals and poorer statistics when using lower PI current for burst BA mode. This effect is most significant for the SrO− profiles. Additionally, we determined the contribution of the gb diffusion following the procedure described by Chung and Wuensch,43 which is based on Whipple’s exact solution. This approach is significantly more accurate than the more commonly used LeClaire approximation 44 for sputter-based analyses, such as ToF-SIMS. For the accurate application of this type of analysis, several conditions must be carefully considered. First, the bulk diffusion distance must be much shorter than the grain size in order to distinguish between the bulk and gb contribution (i.e., Harrison type B kinetics),45 so that the polycrystalline material is not behaving as a homogeneous medium. In the present case, the grain size is about 1 order of magnitude larger than the diffusion length (the average grain size was determined from SEM micrographs as 4.2 ± 1.3 μm), so we can assume that the diffusion takes place under mixed control. Second, this model was established for a constant surface concentration boundary condition. We refer to the previous work on LSM20 by De Souza et al.34 for further details and discussion about these aspects for the specific case of LSM20. Following the analysis of Chung and Wuensch,43 the grain boundary diffusion product (D*gb·δ) is determined by applying eq 5, where A and B are empirical fitting parameters that depend on the slope of the grain boundary tail of the ToFSIMS depth profiles, as tabulated in ref 43.

(3)

18

The experimental O normalized isotopic fraction profiles, and the fits to eq 4 for the SASI (c) and BA (d) modes are also shown in Figure 4. The contribution of oxygen bulk diffusion to the total 18O penetration profile is also included in these plots (dotted blue lines). It should be mentioned that the first C′x values obtained at the beginning of the BA profile (Figure 4d) are slightly lower than those obtained for the SASI profiles (Figure 4c) due to the saturation of the 18O signal for these first points. A closer look to the surface spectra corresponding to these points confirmed the 18O signal saturation after the first burst, and hence, the burst mode should be ignored at the beginning of the BA profile (i.e., the total intensity should by calculated by multiplying the first burst intensity by the total number of bursts). As mentioned previously, the high PI current in the SASI profiles allows the simultaneous detection of other signals such metal oxide secondary ions (MO−, such as LaO−, SrO−, MnO−), which typically have much lower SI yields than the oxygen isotopes (inset plots in Figure 5). In contrast to the BA

B⎤ ⎡ ⎛ ∂lnC(x , t ) ⎞ ⎥ 3/2 1/2 ⎢ A * * ⎟ Dgb ·δ = 2·(D b ) ·t · 10 ⎜ − ⎢⎣ ⎝ ∂η6/5 ⎠ ⎥⎦

Figure 5. Normalized 18O isotopic fraction profiles (C′x) as calculated from the MO− profiles in (a) SASI mode and (b) BA mode. Inset plots correspond to the Poisson-corrected M16O− profiles, with M being the most abundant cation isotope.

A normalized depth x η= D b*·t

mode (inset Figure 5b), the MO− ion signals recorded in the SASI profiles (inset Figure 5a) acquired with high PI current are still of enough intensity to be used reliably for the determination of the oxygen isotope diffusion profile (Figure 5a). This is possible due to the fact that the attenuation is specifically applied to the oxygen SI signals, rather than globally reducing the global intensity of the whole spectrum. The MO− signals in the SASI profile show stable intensities with minimal scatter, almost 30 times higher than those obtained for the PI current-reduced BA mode. In principle, this also allows the use of the MO− oxide signals to determine the isotope diffusion profile, as follows: Cx =

I(M18O) I(M18O) + I(M16O)

(5)

(6)

is defined taking into account the exchange time (t) and bulk diffusion coefficient (Db*), and the gradient of a plot of ln(Cx,t) against η6/5 is taken. The depth range 6 ≤ η ≤ 10 is used to determine this gradient, assuring that the contribution from oxygen diffusion is minimal. These penetration depth ranges (6 ≤ η ≤ 10) used to obtain the grain boundary product as well as the corresponding fitting from the BA and SASI profiles are shown in Figure 6. As observed, this bulk diffusion contribution to the 18O isotopic fraction profiles are the same for both modes (BA and SASI), although the low PI currents used in the BA analyses led to worse signal statistics at the end of the profile when the gb diffusion will be the predominant phenomenon taking place. Therefore, it is clear that the SASI profile provided a better regression coefficient that the BA mode for η6/5 > 6 due to the better 18O limit of detection and, hence, better signal statistics.

(4)

from which reasonable diffusion data may be extracted while simultaneously studying cation segregation and structural F

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry

Souza et al.’s work, the region of the grain boundary tail used for the fitting was shorter than the desirable penetration depth range for this type of analysis, due to the large dispersion of the experimental points for η6/5 > 11 (Figure 7a). The slope in this

Figure 7. (a) 18O penetration profile obtained for the LSM20 sample exchanged at 1000 °C and PO2∼ 1 bar, as measured by De Souza et al. using quadrupole SIMS. Adapted with permission from ref 34. Copyright 2000 Elsevier. (b) Oxygen tracer grain boundary diffusion product as a function of temperature as estimated using quadrupole SIMS and SASI at 1000 °C.

Figure 6. Grain boundary contribution to the diffusion profile based on Whipple’s exact solution to estimate the grain boundary diffusion product, as described by Chung and Wuensch. Residuals of the fitting for the grain boundary region are shown below for the normalized 18O isotopic fraction profiles (C′x) obtained using (a) SASI mode and (b) BA mode.

region led to a value of 5.8 × 10−17 cm3·s−1 for the gb diffusion product (D*gb·δ) at 1000 °C, 1 order of magnitude lower than estimated from our ToF-SIMS analysis. In fact, the gb product value obtained at 1000 °C by De Souza et al. was clearly an outlier in the D*gb· δ vs reciprocal temperature Arrhenius plot (Figure 7b), which was used to estimate the activation energy (Eagb) of grain boundary diffusion in LSM20. The activation energy estimated using the values for temperatures between 700 and 900 °C was Eagb = 317 ± 4 kJmol−1. The improved accuracy of the SASI methodology used here yielded a value of D*gb· δ at 1000 °C, which is consistent with the rest of the data presented by De Souza et al.34 Replacing the data point at 1000 °C with our value yields an activation energy Eagb = 300 ± 8 kJ mol−1, which is only slightly lower than De Souza’s estimated value. Although a re-evaluation and comparison of the oxygen transport behavior of LSM20 is out of the scope of this study, we note the calculated D*gb·δ is consistent with the values from De Souza et al.’s study.34 Although the value of the bulk diffusivity would be expected to be the same for different samples of the same material, this is not necessarily the case for the grain boundary diffusion product. The gb diffusion product contains two terms, the gb diffusivity, D*gb, and the thickness, δ, of the gb “phase” (the analysis treats the grain boundary mathematically as a slab of secondary phase with finite width). Both of these terms may be expected to be sample dependent; as noted in ref 19, different grain boundaries even in the same sample (Al2O3 in that case) show different gb diffusivities. For example, samples may show a different distributions of grain

This improvement will be significant when we have to distinguish between bulk, gb, and dislocation diffusion, which give profiles with similar slopes in this region (e.g., η6/5 dependency for gb or pipe diffusion vs η dependency for diffusion through dislocations). The 18O surface exchange and bulk and gb diffusion coefficients as estimated from the BA and SASI measurements are summarized in Table 2. The relative contribution of the grain boundary and bulk diffusion to the profiles may be characterized using a parameter, β, defined as β=

* ·δ Dgb 3/2

2·D b* ·t 1/2

(7)

In both modes (BA and SASI), we obtain β factors higher than 10, and hence, we can conclude that the experiment was performed under suitable conditions to study the grain boundary diffusivity.44 Similar values were found for oxygen diffusion parameters for the two different modes, although better signal statistics were obtained with the high current SASI mode. Furthermore, as mentioned previously, an additional advantage of the SASI attenuation is the higher signal obtained for other SI species of interest in the samples (e.g., MO− species). More interestingly, the comparison of these results with those reported by De Souza et al., revealed that D*gb·δ obtained for this temperature (1000 °C) was underestimated.34 In De

Table 2. Comparison of the Surface Exchange and Diffusion Coefficients (Grain Boundary and Bulk) Obtained for a Polycrystalline LSM20 Sample Exchanged in 1 bar of 18O2 at 1000°C, As Reported by De Souza et al. (Ref 34) and in This Work method

k* (10−7 cm·s−1)

Db* (10−13 cm2·s−1)

D*gb·δ (10−16 cm3·s−1)

β

10LD (μm)

BA SASI quad. SIMS

0.34 0.39 1.08

5.52 5.71 6.60

5.05 4.62 0.58

14.5 12.6 1.5

6.3 6.4 5.7

G

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Analytical Chemistry boundary types, depending on the degree to preferred orientation, presumably with different effective thickness values, δ, and conceivably different D*gb values. On the other hand, the composition of the grain boundaries may also vary due to the segregation of dopants and impurities to the grain boundaries, which would be very dependent on the impurity content of a given sample. This may also affect D*gb or δ. However, we obtain values in good agreement with those previously published.



ACKNOWLEDGMENTS



REFERENCES

The authors gratefully acknowledge support from the International Institute for Carbon Neutral Energy Research (WPII2CNER), sponsored by the World Premier International Research Center Initiative (WPI), MEXT, Japan. H.T. thanks the financial support from the Japanese Society for Promotion of Science (JSPS postdoctoral fellowship P13770) for her postdoctoral fellowship and the Kakenhi Grant-in-Aid project (25-03770). We would also like to thank Dr. Richard Chater of Imperial College London for many inspiring discussions over the years.



CONCLUSIONS The measurement of oxygen isotopic fractions using ToF-SIMS is often hampered by dead-time effects and detector saturation. In this paper, we have shown the application of a new approach based on Selective Attenuation of Secondary Ions (SASI) to determine oxygen isotope depth profiles with a high dynamic range using ToF-SIMS. To validate the technique, we analyzed the parallel oxygen transport pathways in a perovskitestructured mixed conducting oxide La0.8Sr0.2MnO3. Because this method does not entail a decrease of the primary ion intensity, the SASI approach overcomes the problems associated with high secondary ion intensities when analyzing different isotopes of matrix species while allowing the simultaneous detection of low intensity secondary ions. In this way, the effective dynamic range of the measurement is increased, allowing the determination of 18O isotopic fractions with improved precision and accuracy compared to other methods based on the reduction of the primary ion current. In addition to the accurate determination of the kinetic parameters governing oxygen isotopic exchange, the selfdiffusion coefficient D* and the surface exchange coefficient k*, we were able to determine the diffusion parameters for the grain boundary tail, difficult in other modes of analysis (such as the BA mode) because of low secondary ion intensities of the minority isotope. The improved statistics in the tail region of the profile allow the slope of the tail to be determined more precisely, enabling differentiation between different fast diffusion pathways such as grain boundary and pipe dislocation. Although the technique has been demonstrated for oxygen isotopic analysis in electroceramic materials, it is widely applicable to other matrices bringing additional advantages such as better signal-to-noise ratios (specially for low SI yield species), fast acquisition of images (higher statistics), and simultaneous measurement of other species of interest, which might be important for samples with a limited available analysis area.



Article

(1) Maier, J. Solid State Ionics 1998, 112, 197−228. (2) Ishigaki, T.; Yamauchi, S.; Kishio, K.; Mizusaki, J.; Fueki, K. J. Solid State Chem. 1988, 73, 179−187. (3) Chater, R. J.; Carter, S.; Kilner, J. A.; Steele, B. C. H. Solid State Ionics 1992, 53−56 (Part 2), 859−867. (4) Kilner, J. A.; Steele, B. C. H.; Ilkov, L. Solid State Ionics 1984, 12, 89−97. (5) Choudhury, A.; Palmer, D. W.; Amsel, G.; Curien, H.; Baruch, P. Solid State Commun. 1965, 3, 119−122. (6) Heitjans, P. Solid State Ionics 1986, 18−19 (Part 1), 50−64. (7) Heitjans, P.; Indris, S. J. Phys.: Condens. Matter 2003, 15, R1257− R1289. (8) Kim, B.-K.; Park, S.-J.; Hamaguchi, H.-o. J. Am. Ceram. Soc. 1993, 76, 2119−2122. (9) Stender, D.; Heiroth, S.; Lippert, T.; Wokaun, A. Solid State Ionics 2013, 253, 185−188. (10) Bouwmeester, H. J. M.; Song, C. L.; Zhu, J. J.; Yi, J. X.; Annaland, M. V.; Boukamp, B. A. Phys. Chem. Chem. Phys. 2009, 11, 9640−9643. (11) Kan, C. C.; Kan, H. H.; Van Assche, F. M.; Armstrong, E. N.; Wachsman, E. D. J. Electrochem. Soc. 2008, 155, B985−B993. (12) De Souza, R. A.; Chater, R. J. Solid State Ionics 2005, 176, 1915− 1920. (13) Kilner, J. A.; DeSouza, R. A. In High Temperature Electrochemistry: Ceramics and Metals; Poulsen, F. W., Bonanos, N., Linderoth, S., Mogensen, M., ZachauChristiansen, B., Eds.; 1996; pp 41−54. (14) Fielitz, P.; Borchardt, G. Solid State Ionics 2001, 144, 71−80. (15) Kilner, J. A.; Skinner, S. J.; Brongersma, H. H. J. Solid State Electrochem. 2011, 15, 861−876. (16) Fahey, A. J.; Messenger, S. Int. J. Mass Spectrom. 2001, 208, 227−242. (17) Coker, E. N.; Ohlhausen, J. A.; Ambrosini, A.; Miller, J. E. J. Mater. Chem. 2012, 22, 6726−6732. (18) Holzweber, M.; Kriegl, M.; Schintlmeister, A.; Paesold, D.; Danninger, H.; Hutter, H. Anal. Bioanal. Chem. 2011, 400, 659−663. (19) Heuer, A. H.; Nakagawa, T.; Azar, M. Z.; Hovis, D. B.; Smialek, J. L.; Gleeson, B.; Hine, N. D. M.; Guhl, H.; Lee, H. S.; Tangney, P.; Foulkes, W. M. C.; Finnis, M. W. Acta Mater. 2013, 61, 6670−6683. (20) Claus, J.; Borchardt, G.; Weber, S.; Hiver, J. M.; Scherrer, S. Mater. Sci. Eng. B 1996, 38, 251−257. (21) Fielitz, P.; Borchardt, G.; Schmücker, M.; Schneider, H.; Willich, P. Appl. Surf. Sci. 2003, 203−204, 639−643. (22) De Souza, R. A.; Martin, M. MRS Bull. 2009, 34, 907−914. (23) Opitz, A. K.; Schintlmeister, A.; Hutter, H.; Fleig, J. Phys. Chem. Chem. Phys. 2010, 12, 12734−12745. (24) Kubicek, M.; Rupp, G. M.; Huber, S.; Penn, A.; Opitz, A. K.; Bernardi, J.; Stoger-Pollach, M.; Hutter, H.; Fleig, J. Phys. Chem. Chem. Phys. 2014, 16, 2715−2726. (25) Thoreton, V.; Hu, Y.; Pirovano, C.; Capoen, E.; Nuns, N.; Mamede, A. S.; Dezanneau, G.; Yoo, C. Y.; Bouwmeester, H. J. M.; Vannier, R. N. J. Mater. Chem. A 2014, 2, 19717−19725.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Present Address ∥

School of Chemical Engineering, University of Birmingham, Edgbaston, Birmingham, B15 2TT, United Kingdom. Author Contributions ⊥

H.T. and J.D. contributed equally. The manuscript was written through contributions of all authors. All authors have given approval to the final version of the manuscript. Notes

The authors declare no competing financial interest. H

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX

Article

Analytical Chemistry (26) Tellez, H.; Aguadero, A.; Druce, J.; Burriel, M.; Fearn, S.; Ishihara, T.; McPhail, D. S.; Kilner, J. A. J. Anal. At. Spectrom. 2014, 29, 1361. (27) Zapata, J.; Burriel, M.; Garcia, P.; Kilner, J. A.; Santiso, J. J. Mater. Chem. A 2013, 1, 7408−7414. (28) Hancke, R.; Fearn, S.; Kilner, J. A.; Haugsrud, R. Phys. Chem. Chem. Phys. 2012, 14, 13971−13978. (29) Langer-Hansel, K.; Schintlmeister, A.; Fleig, J.; Hutter, H. Surf. Interface Anal. 2013, 45, 486−489. (30) Stephan, T.; Zehnpfenning, J.; Benninghoven, A. J. Vac Sci. Technol. A 1994, 12, 405−410. (31) De Souza, R. A.; Zehnpfenning, J.; Martin, M.; Maier, J. Solid State Ionics 2005, 176, 1465−1471. (32) Holzlechner, G.; Kubicek, M.; Hutter, H.; Fleig, J. J. Anal. At. Spectrom. 2013, 28, 1080−1089. (33) Kubicek, M.; Holzlechner, G.; Opitz, A. K.; Larisegger, S.; Hutter, H.; Fleig, J. Appl. Surf. Sci. 2014, 289, 407−416. (34) De Souza, R. A.; Kilner, J. A.; Walker, J. F. Mater. Lett. 2000, 43, 43−52. (35) Niehuis, E. U.S. Patent No. US 8785844B8785842, 2014. (36) IonToF. ToF-SIMS V technical information. Münster, Germany, 2014. (37) Böhlke, J. K.; De Laeter, J. R.; De Bièvre, P.; Hidaka, H.; Peiser, H. S.; Rosman, K. J. R.; Taylor, P. D. P. J. Phys. Chem. Ref. Data 2005, 34, 57−67. (38) van Heuveln, F. H.; Bouwmeester, H. J. M.; van Berkel, F. P. F. J. Electrochem. Soc. 1997, 144, 126−133. (39) Jiang, S. J. Mater. Sci. 2008, 43, 6799−6833. (40) Crank, J. The Mathematics of Diffusion; Oxford University Press: Oxford, U.K., 1975. (41) Szot, K.; Speier, W.; Herion, J.; Freiburg, C. Appl. Phys. A: Mater. Sci. Process. 1996, 64, 55−59. (42) Druce, J.; Téllez, H.; Burriel, M.; Sharp, M.; Fawcett, L.; Cook, S. N.; McPhail, D.; Ishihara, T.; Brongersma, H. H.; Kilner, J. A. Energy Environ. Sci. 2014, 7, 3593−3599. (43) Chung, Y. C.; Wuensch, B. J. J. Appl. Phys. 1996, 79, 8323− 8329. (44) Leclaire, A. D. Br. J. Appl. Phys. 1963, 14, 351−356. (45) Harrison, L. G. Trans. Faraday Soc. 1961, 57, 1191−1199.

I

DOI: 10.1021/ac504409x Anal. Chem. XXXX, XXX, XXX−XXX