Anal. Chem. 2006, 78, 7273-7277
Determination of the Bulk Cobalt Valence State of Co-Perovskites Containing Surface-Adsorbed Impurities O. Haas,*,† Chr. Ludwig,†,‡ and A. Wokaun†
General Energy Research, Paul Scherrer Institut, CH-5232 Villigen PSI, Switzerland, and Swiss Federal Institute of Technology, ENAC-ISTE, CH-1015-Lausanne, Switzerland.
We used thermogravimetric hydrogen reduction and iodometric titration to determine the bulk valence state of cobalt in Co-perovskites containing surface carbonate hydroxide or hydroxyl groups. It could be shown that thermogravimetric hydrogen reduction experiments are very sensitive to volatile surface groups, but due to their volatility, they can be specified and the bulk valence state of cobalt can still be deduced from these experiments. The iodometric titration is less sensitive to small volatile surface impurities, but precaution has to be taken that oxygen or iodine does not escape from the solution during dissolution of the sample. Best results were obtained if the sample was titrated during dissolution in a closed argon floated titration apparatus. We tested the two methods using LaCoO3 perovskite as a sample with a known valence state. Both methods delivered satisfactory results, and the valence state could be determined with an accuracy of better than 1%. Cobalt oxides with a perovskite structure are endowed with interesting catalytic, ion conduction, magnetic, and semiconducting properties. LaCoO3 exhibits catalytic activity above 700 °C, where CO can be oxidized and NOx decomposed.1 La1-xCaxCoO3-δ can be used in bifunctional porous oxygen diffusion electrodes as a catalyst to reduce and to evolve oxygen,2,3 while La1-xSrxCoO3-δ is widely used as an oxygen ion-conducting and electron-conducting electrode material in high-temperature fuel cells.4 The electronic and magnetic properties of holes-doped cobalt perovskites have also been of paramount interest in the last three decades. Depending on the application, surface or bulk properties are important for the desired chemical and physical properties. In an earlier study, we investigated La1-xCaxCoO3-δ perovskites as a catalyst in oxygen diffusion electrodes5 and X-ray absorption spectroscopy was used to characterize the electronic and molec* Corresponding author. Fax: 0041 56 310 2199. E-mail:
[email protected]. † Paul Scherrer Institut. ‡ Swiss Federal Institute of Technology. (1) Voorhoeve, R. J. H. Advanced Materials in Catalysis; Academic Press: New York, 1977. Libby, W. F. Science 1971, 171, 499. (2) Mu ¨ ller, S.; Striebel, K. A.; Haas, O. Electrochim. Acta 1994, 39, 1661. (3) Bursell, M.; Pirjamali, M.; Kiros, Y. Electrochim. Acta 2002, 47, 1651. (4) M. H. R. Bouwmeester Verweij, H. Solid State Ionics 1997, 96, 21. (5) Mu ¨ ller, S.; Holzer, F.; Arai, H.; Haas, O. J. New Mater. Electrochem. Syst. 1999, 2, 227. 10.1021/ac060903w CCC: $33.50 Published on Web 09/12/2006
© 2006 American Chemical Society
ular properties of these compounds.6,7 XPS and soft XAS measurements using electron yield detection are surface-sensitive analytical tools. In contrast, XAS investigations of the Co K-edge give a superimposed signal of the bulk and surface properties, which at normal grain size is dominated by the bulk properties. For a correct interpretation of the XAS spectra, it is very helpful to have additional information about the stoichiometry and the valence state of the cobalt in the bulk of the perovskite. The surfaces of Co-perovskites are alkaline and have the tendency to absorb CO2 and water molecules out of the air. If the perovskites are stored in air, the perovskite will be covered with at least a monolayer of these molecules. In some cases, this weathering reaction may go on. These surface impurities depend on the specific surface area and are normally in the range of 0.5-3 wt %. At the surface of the perovskite, a very small part of the perovskite may change its structure and the involved transition metal might even change its valence state, whereas the bulk properties remain unchanged. For many analytical investigations, these surface impurities will not show up; e.g., transmission X-ray absorption measurements at the transition metal K-edge or X-ray diffraction pattern will not be influenced markedly. In technical applications where the bulk properties are important, these surface impurities may not be of interest or any harm. However, if the bulk valence state of these perovskites has to be evaluated, even small surface impurities may cause an analytical problem. Depending on the method used, the result may be falsified considerably. We tested thermogravimetric hydrogen reduction and iodometry to evaluate the bulk valence state of air-stored cobalt perovskites and used LaCoO3 as a sample for these tests. The valence state of cobalt in LaCoO3 prepared in air at 750 °C is expected to be 3+ due to the fact that the valence state of La in such oxides is always 3+ and the one of oxygen 2-. Oxygen vacancies (δ value) can be neglected if the sample is prepared at ambient oxygen pressure. In fact, oxygen vacancies are only expected if a very low partial oxygen pressure is used during preparation.8 LaCoO3 is, therefore, a good reference material to (6) Haas, O.; Holzer, F.; Mu ¨ ller, S.; McBreen, J. M.;. Yang, X. Q.; Sun, X.; Balasubramanian, M. Electrochim. Acta 2002, 47, 3211. (7) Haas, O.; Struis, R. P. W. J.; McBreen, J. M. J. Solid State Chem. 2004, 177, 1000. (8) Nakamura, T.; Petzow, G.; Gauckler, L. J. Mater. Res. Bull. 1979, 4, 649.
Analytical Chemistry, Vol. 78, No. 20, October 15, 2006 7273
Figure 2. TG curve of LaCoO3 obtained during hydrogen reduction. The second reduction was started after reoxidation of the sample in air at 750 °C. Figure 1. Apparatus used for the iodometric titration. (1) 0.05 M Na2S2O3 solution, (2) step motor, (3) magnetic stirrer, (4) perovskite supply, (5) reference electrode, and (6) Pt electrode.
test analytical methods designed for valence state determination of cobalt in Co-perovskites. Thermogravimetric hydrogen reduction has been employed rather frequently9-11 to determine the valence state or oxygen vacancy (δ) of cobalt oxides. The same holds for the iodometric titration.9,12,13 However, in many cases, we think that the determined δ values or the valence state of cobalt is not as accurate as indicated or even wrong. Both methods have intrinsic problems, which are addressed in this paper. EXPERIMENTAL SECTION Synthesis of LaCoO3. LaCoO3 perovskite was prepared using the citrate acid precursor method.3 Stoichiometric amounts of La(NO3)3‚6H2O (Fluka) and Co(NO3)2‚6H2O (Fluka) were added to an aqueous 1 M citric acid solution. The solution was evaporated at 70 °C using a rotary evaporator and dried for 3 days at 90 °C. The resulting powder was first heated in air at 200 °C for 2 h and then calcined in air for another 2 h at 700-750 °C. The samples were characterized using XRD, BET, and XPS. The XRD data were obtained using a Philips X-Pert diffractometer with Cu KR radiation. The BET surface area was measured using the surface area analyzer Micromeritic ASAP 2000. The XPS measurements were measured using the XPS analyzer from VG Scientific. Thermogravimetric Experiments. A. Mettler-Toledo TGA/ SDTA851e equipment was used for the thermogravimetric experiments. The temperature scans were performed under reducing (5% H2 in Ar) and oxidizing (21% O2 in Ar) conditions. The gas flow was adjusted to 50 mL/min, and the heating rate was set to 3 °C/min. Iodometric Titration and Data Treatment. Figure 1 shows the apparatus used for the iodometric titration. Typically ∼10 mg of perovskite was suspended with a magnetic stirrer under Ar in 10 mL of 4 M HCl containing 0.5 g of KI. The HCl/KI solution (9) Karppinen, M.; Matvejeff, M.; Saloma¨ki, K.; Yamauchi, H. J. Mater. Chem. 2002, 12, 1761. (10) Kasper, N. V.; Troyanchuk, I. O.; Khalyavin, D. D.; Hamad, N.; Haupt, L.; Fro¨bel, P.; Ba¨rner, K.; Gmelin, E.; Huang, Q.; Lynn, J. W. Phys. Status Solidi B 1999, 215, 697. (11) Kim, W. S.; Chi, E. O.; Choi, H. S.; Hur, N. H.; Oh, S.-J.; Ri, H.-C. Solid State Commun. 2000, 116, 609. (12) Irvine, J. T. S.; Namgung, C. J. Solid State Chem. 1990, 87, 29. (13) Schweizer, T. Diss. ETH Zu ¨ rich, No. 10167, 1993.
7274
Analytical Chemistry, Vol. 78, No. 20, October 15, 2006
was first prepared under argon before the perovskite was added. The solution was then slowly titrated (typically with of 1500 s/mL) under argon using a Metrom titrimat and a 0.05 M Na2S2O3 solution. A Pt electrode was used to sense the redox potential of the solution and an Ag/AgCl electrode as a reference electrode. The equivalent point was evaluated taking the highest gradient at the end of the titration curve. This value was corrected by subtracting 0.021 mL for the I2 content of the HCl/KI solution and by the concentration factor of the Na2S2O3 solution determined using KIO3 as a standard. The 0.021 mL was evaluated by titrating a similar prepared solution containing no perovskite. To evaluate the valence state of cobalt in the bulk a correction for the volatile surface impurity was also necessary. We accounted for that by subtracting 1.6% of the initial weight. The weight loss of 1.6% was evaluated from the data shown in Figure 2. It was calculated from the difference of the initial weight between the first and second reduction scan. We made no correction for a possible change in valence state of the outermost cobalt layer, assuming that this layer is not changing its valence state by adsorbing CO2 and H2O. RESULTS AND DISCUSSION Synthesis and Characterization of LaCoO3. The LaCoO3 perovskite had typically a surface area of 10-20 m2/g. The phase purity was checked using XRD; it revealed a single phase with no indication of an impurity. The XRD was in agreement with the data published by Thronton et al.14 listed in the ICSD,15 where the pattern could be indexed using the rhombohedral space group R-3c, No. 167. Thermogravimetric Hydrogen Reduction of LaCoO3. The thermogravimetric hydrogen reduction method makes use of the fact that all cobalt oxides can be reduced at elevated temperature to metallic cobalt if a reaction medium of 5% hydrogen in 95% Ar is used. The reaction going on during the thermogravimetric reduction of LaCoO3 can be described as follows:
LaCoO3 + 1.5H2 f 0.5 La2O3 + Co + 1.5H2O
(1)
Only the cobalt is reduced, while La(III) is stable in the temperature range used in our experiments. The weight difference produced by the reduction process is small compared to the total weight of the probe (10.82% of the original LaCoO3 weight). (14) Thronton, G.; Tofield, B. C.; Hewat, A. W. J. Solid State Chem. 1986, 61, 301. (15) ICSD for WWW Inorganic Crystal Structure Database.
Figure 2 (first reduction) presents the weight loss curve obtained during hydrogen reduction of the air-stored LaCoO3 sample. It shows the weight loss at the right scale and the corresponding stoichiometry of the reduction process at the left scale. The stoichiometry factor at the left side is indicative of the change in valence state during the reduction process. It was calculated using the initial weight, the total weight loss, and the theoretical weight loss expected due to eq 1. The samples were scanned up to 750 °C. At 750 °C, a very flat plateau indicates the completeness of the expected reduction process. The starting weight of the sample in the experiment shown in Figure 2 at 100 °C was 42.673 mg. From the stoichiometry of the reaction, the theoretical weight loss is expected to be 4.166 mg. The experimental weight loss, however, was 4.845 mg. The obtained weight loss indicates a valence state that is significantly higher than 3+. In fact, from the weight loss curve we deduce a valence state of 3.49+, which is 16% more than expected. The sample is losing 0.679 mg more than expected, which corresponds to 1.6% of the total weight. This additional weight loss is reasonably explained with a corresponding amount of volatile impurities. As stated in the introduction, perovskites such as other metal oxides have an alkaline surface, which can adsorb CO2 and water molecules leading to carbonates, hydroxide, or hydroxyl groups at the surface. These two weathering reactions are expected to be important, especially if the specific surface area of the perovskite powder is high. The mentioned surface groups are volatile under the reaction condition used in our thermogravimetric hydrogen reduction experiment. We think the surface of our sample is mainly covered with carbonates and bicarbonates since primary hydroxy groups should be able to react with CO2 to form carbonates or bicarbonates. This hypothesis was also verified with a preliminary XPS investigation. The XPS measurements showed a relative strong C1s and O1s signals corresponding to about 22% carbon and 52% oxygen in the surface layer accessible by XPS. This is most probably due to surface carbonates and bicarbonates.16 However, the XPS result should not be overinterpreted since surface carbon is often observed if XPS is used to analyze untreated powder samples. An investigation of air-stored La0.6Ca0.4CoO3-δ samples using TGA in connection with FT-IR spectroscopy17 revealed CO2 and H2O escaping from these samples during the reduction scan. Although these measurements were not done on the discussed LaCoO3 sample, it can be regarded as a support of the hypothesis that perovskites form carbonates or bicarbonates at the surface if stored in air. In a recent investigation Islam18 showed that the LaCoO3 is exposing mainly {100} and {110} surfaces, where the {100} surface terminates with lanthanum (A-site) and oxygen ions and the {110} surface with an oxide layer. We estimate that the adsorbed surface groups occupy about 10-15 Å2. A specific surface area of 15 m2/g and a surface area of about 10-15 Å2 for each CO2 adsorbed would lead to a volatile impurity close to 1% and pretend a valence state that is ∼10% too high. This calculation may be off by 50%, but it certainly demonstrates the sensitivity of the TGA measurements to volatile surface impurities. It reveals (16) Personal communication, Schnyder, B. Paul Scherrer Institut, 2003. (17) Unpublished results: Haas, O.; Ludwig, Ch.; Struis, R. P. W. J. (18) Islam, M. S. Solid State Ionics 2002, 75, 154-155.
that our sample had a coverage somewhat higher but still in the order of a monolayer. If we carefully analyze the TGA curve, we recognize that the volatile impurity is only influencing the first step. The weight reductions in the second and third steps are equal and as theoretically expected. That means that all the volatile impurity is escaping before the second and third steps. With an additional experiment on an untreated sample, we could show that the surface carbonates and hydroxyls can also be removed in a heating cycle using a gas mixture containing Ar + 21% O2. After the first reduction scan, the sample was reoxidized with a temperature scan up to 750 °C in a gas mixture containing Ar + 21% O2. At this temperature and partial oxygen pressure, the cobalt is oxidized to cobalt oxides and the structure of the LaCoO3 perovskite is finally re-formed. This can be expected since these are normal conditions to synthesize the perovskite out of wellmixed oxides. The procedure allows the reproduction of the perovskite under conditions that do not allow the formation of surface carbonates and hydroxyls. After the perovskite was reformed, the sample was cooled under oxidizing conditions and a second reduction scan using again 5% H2 in Ar was started. The thermogravimetric curve obtained from this second reduction scan is shown in Figure 2 (second reduction). The close match of the first and second reduction curves in the second and third reduction steps is a good indication for the re-formation of the perovskite during the oxidizing scan. The total weight loss in this second reduction scan corresponds almost exactly to the value expected for a three-electron reduction. It is interesting to see that the reduction conditions used allow the production of all three valence states of the cobalt. The straight line to start with and the step size of the first step in this curve (equal to the second and third steps) indicates that no volatile impurities are present anymore. With an additional experiment on an untreated sample, we could show that the surface carbonates and hydroxyls can also be removed in a heating cycle using a gas mixture containing Ar + 21% O2. Taking the new initial weight equal to 41.8724 mg, we calculate a theoretical final weight of 37.7845 mg but obtain an experimental final weight at 750 °C of 37.7637 mg. From these numbers we evaluate a valence state of 3.015 and thus a δ value of -0.0075, which is within 0.5% close to the theoretical value and within the accuracy (2.6 before the solution is titrated with Na2S2O3. This makes the titration less sensitive to air, but the neutralization has to be performed under an inert gas. We modified the standard iodometric titration in the sense that we made the titration during the dissolution process in the original 4 M HCl solution under rigorous oxygen exclusion. This procedure has the advantage that the dissolution process becomes much faster. This is due to the fact that cobalt(II) oxides are dissolving much faster than cobalt(III) oxides. It could be shown that intrinsic first-order dissolution constants are directly related to the water exchange rates of ion hydrate complexes in solution,19 where the exchange rates are known to be much faster for lower valences. The dissolution rates of oxides are, therefore, strongly influenced by the valency of the metal ions in the crystal structure,20 and Co(II) compounds are for this reason expected to dissolve much faster than Co(III) compounds. If the perovskite is titrated during dissolution, we can expect that Co(III) is reduced to Co(II) at least at the surface of the perovskite crystals, which then accelerates the dissolution process. (19) Casey, W. H.; Ludwig, Chr. Rev. Mineral. 1995, 31, 87. (20) Hubli, R. C.; Mittra, J.; Suri, A. K. Hydrometallurgy 1997, 44, 125.
7276
Analytical Chemistry, Vol. 78, No. 20, October 15, 2006
Figure 4. Iodometric titration of 11.37 mg of LaCoO3 in 4 M HCl using a 0.05 M Na2S2O3 solution. Table 1. Results of the Iodometric Titrationsa sample wt (mg)
correction for adsorbed CO2
mL of Na2S2O3 used
corrected mL of 0.05 M Na2S2O3
calcd valence state
11.37 11.81 12.06
11.19 11.62 11.87
0.890 0.940 0.949
0.908 0.959 0.968
2.992 3.038 3.007
a Molecular weight of LaCoO , 245.85. Average valence state: 3.01 3 ( 0.025 f LaCoO3.005.
In addition, the iodine and oxygen concentration is kept very low during the dissolution process, which reduces the risk that I2 or O2 can escape before it is titrated with Na2S2O3. Using this procedure, the titration including the dissolution took ∼1/2 h. To control whether all of the material was dissolved at the end point, the titration was stopped shortly after the equivalence point to see if it remained constant. If the dissolution would go on, the potential would increase. Using this method we were able to determine the correct valence state of Co in the perovskite. Figure 4 shows the titration curve obtained from a titration of 11.37 mg of LaCoO3. At the beginning of the titration, the potential is increasing due to the dissolution of Co(III). The potential is then decreasing due to the production of Co(II) induced by the titration agent Na2S2O3. The stoichiometry of the titration reaction is as follows:
Co3+ + I- f Co2+ + 1/2I2 1
/2I2 + S2O32- f I- + 1/2S4O62-
The evaluation of the valence state from the titration curve is explained in detail in the Experimental Section. The results of three titrations are presented in Table 1. We obtain a valence state of 3.01 ( 0.025 or a δ-value of -0.005 ( 0.01. All three titrations listed in Table 1 are within the expected accuracy (∼1%) of the iodometric titration. CONCLUSIONS For the determination of the valence state in Co-perovskites, the thermogravimetric hydrogen reduction method is only reliable if volatile impurities are rigorously excluded from the sample. Perovskites such as LaCoO3 have the tendency to adsorb CO2 at the surface, which can lead to an overestimation of the valence state. In the case of LaCoO3, only the first reduction step is
influenced by this additional weight loss. The iodometric titration is less sensitive to these impurities, but much more demanding from an experimental point of view. Valuable results were obtained if the perovskite was titrated during dissolution in a closed argon floated titration apparatus. ACKNOWLEDGMENT We acknowledge A. Schuler for the excellent experimental support concerning the thermogravimetric hydrogen reduction
experiments and the development of the iodometric titration method. We thank Dr. Marcel Sturzenegger, PSI, for critical comments.
Received for review May 17, 2006. Accepted August 9, 2006. AC060903W
Analytical Chemistry, Vol. 78, No. 20, October 15, 2006
7277
