J. MATER. CHEM., 1994,4(5), 663-668 Preparation of an Na,Zr,Si,PO,,-Sodium Aluminosilicate Composite and its Application as a Solid-state Electrochemical C02 Gas Sensor Susumu Nakayama" and Yoshihiko Sadaokab a New Materials Research Center, Shinagawa Refractories Co. Ltd., Bizen, 705 Japan Department of Applied Chemistry, Faculty of Engineering, Ehime University, Matsuyama, 790 Japan A composite of sodium aluminosilicate and Na,Zr,Si,PO1, has been examined as a dense solid-state sodium-ion conductor. For the composite sintered at 1050 "C, the electrical properties are mainly based on those of Na3Zr2Si2PO12. When the composite was sintered at a higher temperature, the activation energy of the resistance increased to ca. 55 kJ mol-' and the densification was well progressed with the formation of small zirconia particles.A good junction between the composite and Y-stabilized zirconia was achieved when the sintering was carried out at 31175 "C. The formation of zirconia particles in the composite layer resulted in an enhancement of the mechanical strength at the junction. By using the ceramics, a solid-state electrochemical CO, sensor with an Na,CO, layer was fabricated. Whilst the sensitivity of this sensor to CO, was slightly influenced by the coexistence of the water vapour in the test gas, a good sensing characteristic based on two-electron electrochemical reaction was confirmed. The discovery of Na3Zr,Si,PO12 represented an important development in the field of solid electrolytes because it demon- strated that a three-dimensional framework structure could have a conductivity comparable to that of b-alumina, a two- dimensional network.Recently there has been interest in solid glass electrolyte^.'-^ Glasses present several advantages over crystalline materials for solid-state electrolyte and/or chemical sensor application^.^.^ Hunter and Ingram examined sodium- ion conduction in glasses, including silicates and borates, and found that the conductivity increases with optical basicity, whilst the activation energy falls towards an apparent limiting value of ca. 50 kJ mol-1.8 Similar results were reported by Hakim and Uhlmann.g The ceramics obtained by the sintering of a mixture of Na,Zr,Si,PO,, and sodium aluminosilicate indicated that the high ionic conductivity and its activation energy was lowered to ca.50 kJ mol-' and the densification was well pr~gressed.~ Both materials are superior in respect of water resistivity. This paper presents the results of a study of using the title composite to prepare a dense ionic conductor for a solid-state electrochemcial CO, gas sensor. Experimental Crystalline Na,Zr,Si,PO,, powder was made from reagent- grade Na,CO,, NH4H,P04, ZrO, and SiO, using conven- tional ceramic techniques. The mixture of raw materials was ball-milled with ethanol, dried at 100 "C, calcined in air for 4 h at 900 "C, then ground. Pellets were pressed at 1000 kg cmP2 and sintered at 1250 "C in air for 15 h. Na,Zr,Si,PO,, powder was prepared by milling the prepared ceramics (mean particle diameter z1 pm).Na,O-Al,03-4Si02 was made from reagent-grade Na,C03, Al,03 and SiO, by sintering a mixture at 1350 "C. The powders were also obtained by milling. A mixture of Na,Zr,Si,PO,, and 40 wt.% Na20-A1,O3-4SiO2 was compressed and sintered at 1000 "C. ZrO, discs with 8 mol% Y203 were also sintered at 1600 "C. The discs were 10 mm in diameter and ca. 1 mm thick. The sodium-ion conductors were placed on the ZrO, discs and sintered at various temperatures. The crystalline phases were identified at room temperature by the standard X-ray diffraction technique (XRD). The microstructures were examined using scanning electron microscopy (SEM). The electrical properties of discs in which platinum paint acted as electrodes (applied to opposite faces by sintering at 800 "C) were measured using an LCZ meter (100 Hz-10 MHz).A disc with porous platinum electrodes was fixed on the top of an alumina tube with an inorganic adhesive containing sodium. The electrode inside the tube acted as the reference electrode. The platinum sensing elec- trode was positioned on the outside face of the disc, covered with Na,C03 and then dried at 80 "C. Standard gases, air (CO<1 ppm, CO, <2 ppm, HCl < 1 ppm and H,O < 10 ppm) or CO, at 10, 100, 1000 and 10000 ppm diluted with air, were introduced to the working (sensing) electrode side. Humidification of the test gas was achieved by allowing the test gas to bubble through water at 30 "C. The emf of the sensor was measured with a digital electrometer.Results and Discussion Form of Electrolytes Some sodium aluminosilicates prepared by calcination at 1000 "C were examined by XRD. Na,0-A1,03-2Si0, was identified as carnegieite (low form) and no distinct glass phases were detected. For Na,O-Al2O,-4SiO,, the broad band caused by the glass phase and some weak diffraction peaks were detected. For Na,O-Al,O,-nSiO, (n=6 and 8), strong peaks assigned to a-quartz were detected at ti =0.4262 and 0.3343 nm with a broad band. In addition, NASICON was identified in its monoclinic form with ZrO, as an impurity. Previously it was confirmed that when the composites were sintered at 1000 "C, a well densified composite of N.4SICON and Na20-A1,03-4Si0, was obtained for a mixlure with 40 wt.% of Na,O-Al,O,-4SiO2.In this work, a NASICON-glass composite was prepared from a mixture of NASICON with 40 wt.% Na,0-A1,03-4Si0,. The XRD parameters are summarized in Table 1. For the composite sintered at 1050 "C, the observed XRD pattern is very similar to NASTCON singly and the peaks observed for the Na20-A1,03-4Si02 were not detected. When the sintering temperature is increased to 1150 "C, peaks assigned to NASICON were observed and the intensity of the peaks assigned to ZrO, (d=0.316, 0.284 and 0.262 nm) increased with increasing sintering temperature. For samples sintered at 31175 "C, the XRD pattern changed completely, i.e. the disappearance of the signals assigned to NASI(10N and growing-in of the ZrO, signals were observed. The correlation between the ratio of the intensity, I(d=0.294 nm)/l(d= 0.316 nm) and the sintering temperature is shown in Fig.1 in J. MATER. CHEM., 1994, VOL. 4 Table 1 Relative intensities of some XRD peaks sintering temperature/"C composite NASICON d/nm 1200 1050 1100 1125 1150 1175 1200 ~ 0.6510 0.55 0.51 0.52 0.53 0.39 0.09 0.4643 0.94 0.91 0.93 0.96 0.64 0.4520 1.oo 0.99 0.99 1.00 0.69 0.3890 0.52 0.53 0.54 0.55 0.42 0.3694 0.11 0.18 0.24 0.26 0.31 0.25 0.26 0.3233 0.68 0.65 0.67 0.70 0.49 0.3166 0.12 0.34 0.57 0.56 1.00 1.00 1.00 0.2932 0.98 1.00 1.00 0.98 0.70 0.2840 0.12 0.29 0.41 0.51 0.68 0.67 0.71 0.2613 0.36 0.32 0.36 0.36 0.33 0.27 0.25 0.2608 0.74 0.74 0.82 0.85 0.69 0.2542 0.07 0.15 0.18 0.21 0.24 0.21 0.21 0.2495 0.07 0.10 0.12 0.11 0.12 0.09 0.10 Fig.1 Correlation between Z(d =0.294 nm)/Z(d=0.316 nm) and the sintering temperature of the mixture of NASICON and 40 wt.% Na20-A120,-4Si02 which the peaks at d=0.294 and 0.316nm are attributed to NASICON and ZrO,, respectively. The changes in the microstructure were also examined by SEM (Fig. 2). For the NASICON singly, monoclinic crystal- lites 1-2 pm in size were detected. Interfusion of the particles was only partial. The surface of the Na,O-Al,03-4Si0, glass prepared by sintering at 1000 "C was very smooth and there were hardly any detectable pores and/or holes. When a mixture of Na20-A1,O3-4SiO2 glass and NASICON was sintered at 1050 "C, the surface was covered by a glass-like phase and no crystalline phase could be directly observed.It seemed that the NASICON particles were covered by the glass phase whilst XRD signals assigned to NASICON were detected, as mentioned above. When the sintering temperature was increased to 1100 "C, the densification by the glass phases was well progressed, with the formation of cubic-like crystals, whilst the presence of the NASICON particles was detected. The size of the newly formed crystals was smaller than that of the NASICON used as the starting material. The newly detected finer crystals were ascribed to ZrO,. With an increase in the sintering temperature, the formation of ZrO, particles and new glass-like phases were accelerated and the NASICON phase disappeared.When the composite was sintered at 211 75 "C, interfusion of the glass phases proceeded and small closed pores of 0.5-1 pm diameter were observed. The results of the XRD and SEM observations confirm that the NASICON layer reacts with the glass layer and the formation of ZrO, proceeds when the temperature is increased to ca. 1100 "C. Furthermore, to obtain a good junction between the com- posite and the Y-stabilized zirconia disc, the composite disc prepared by sintering at 1000 "C was placed on the zirconia disc and calcined at various temperatures. When the tempera- ture was increased to ca. 1125 "C, cracks formed in the composite disc parallel to the connected surface. For the samples heated at 31150 "C cracks were not detected and each layer was well connected; in addition, the composite disc was softened and partly melted when the temperature was increased to 3 1175 "C.Fig. 3 shows photomicrographs of the fractured faces. For samples sintered at d 1125 "C, no new phases were detected at the junction. When the sample was sintered at 1200 "C, a new phase/layer was detected at the junction. We tried to obtain a good junction between NASICON and Y-stabilized zirconia by sintering at 1200 "C, but an adequate junction could not be obtained. Further detailed experiments on the junction are now in progress. For samples calcined at 31175 "C, the junction does not undergo peel-off or removal and is stable to mechanical and/or thermal shock. The observed properties are sufficient for a solid-state electrochemical CO, gas sensor coated with an Na2C03 layer and with platinum electrodes on the surfaces to be fabricated.These changes in the composite and the enhancement of the mechanical strength of the junction are attributed to the formation of a layer of ZrO, and new glass-like phases in the composite layer and the appearance of this new layer at the junction. Electrical Properties For the composites, the equivalent total electrical resistance is made up from several components such as the intergranular, bulk and electrode-electrolyte junctions. For the Na,O-A1203-4Si0, glass prepared by sintering the mixture at 1000 "C, the complex impedance plot is represented by an arc which passes through the origin.In this case the resistance, which can be estimated from the intercept (Rglass)to the Z'-axis in the low-frequency region, is attributed to the glass phase. Similar features in the complex impedance plots were observed for the composite prepared by sintering at 1200 "C in which the resistivity of the glass phase was estimated from the intercept to the 2'-axis at low frequency. The complex impedance plots for NASICON are shown in Fig. 4. At a low temperature, the complex impedance is represented by an arc in the high-frequency region and by a spur in the low- frequency region. The intercepts (denoted A and B in the figure) to the Z-axis indicate the resistances of the bulk component @bulk) and of the sum of the bulk and grain components (Rbulk+ Rgrain),respectively.When the tempera- ture was high, only the spur was observed, in which the intercept (B) to the Z-axis indicates the sum of the bulk and grain components. For the composite sintered at 1050 "C,the complex impedance plot was more complex (Fig. 5), while only a spur was observed at 250 "C. At 80 'C the high- frequency region was represented by an arc and the low- frequency region by the combination of an arc and a spur. When the temperature was increased to 150 C, the result was represented by the combination of an arc and a spur. The intercept (C) to the 2'-axis of the spur is attributed to the total equivalent resistance (Rbul,+ Rgrain+ Rglass). The values estimated from the intercepts (A, B, C) for the composite at 80 "C corresponded to the bulk (Rbu]k), the sum of the bulk and grain components based on NASICON (Rb,]k + Rgrain) and the sum of Rbu1k, Rgrainand the resistance of glass phase J.MATER. CHEM., 1994, VOL. 4 Fig. 2 Scanning electron micrographs: (a) NASICON sintered at 1200 "C, (b)Na20-A120,-4Si02 sintered at 1000 'C, (c) composite sintered at 1050 "C, (d)composite sintered at 1100 "C, (e)composite sintered at 1150 "C, (f)composite sintered at 1175 "C (Rglass),respectively. From the result observed at 150 "C, where R is the resistivity, R, is the pre-exponential factor, E, (Rbulk +Rgrain)and (Rbu&+Rgrain+Rglass)could be estimated. is the activation energy, kB is the Boltzmann constant and T The temperature dependence of the estimated resistivity is is the absolute temperature.shown in Fig. 6. The results were parametrized by the The activation energy of the resistivity estimated from Arrhenius equation: (Rbulk +Rgrain) of the composite sintered at 1050 "C was comparable to that of NASICON alone. Furthermore, the RT-' =R, exp(E,/k,T) (1) activation energy of the resistivity estimated from Kglasswas J. MATER. CHEM., 1994, VOL. 4 Fig. 3 Scanning electron micrographs of the fractured faces of the (a) 1100 "C, (b) 1200 "C 0 0 0 B A E ,?'(arb. units) Fig. 4 Complex impedance plots of NASICON: (a) 50 "C, (b)200 "C A B B cc Z'(arb. units) Fig. 5 Complex impedance plots of the composite sintered at 1050 "C: (a) 80 "C, (b)150 "C composite-Y-stabilized zirconia junction.Sintering temperature: 4r -3 1.o 1.5 2.0 2.5 1O~WT Fig. 6 Temperature dependence of the resistivities: (Rbulk+Rgrain)of NASICON (a), (Rbulk+Rgrain)of composite sintered at 1050 "C (b), Rglass of composite sintered at 1050 "C (c), Rglass of Na,0-A1203-4Si02 sintered at 1000 "C (d), Rglass of composite sintered at 1200 "C (e) comparable to that of the Na20-Al2O3-4Si0, glass prepared by sintering the mixture at 1000 "C and/or of the composite sintered at 1200 "C.These correlations confirmed the results of XRD and SEM observations. The complex impedance plots for the Y-stabilized zirconia are shown in Fig. 7. A spur in a low-frequency region and an arc which passed through the origin in the complex impedance plot were observed at <300 "C.The intercept (A) to the 2'-axis is attributed to the bulk component.At a higher tempera- ture, two arcs and a spur were observed and the arc in the high-frequency region passed through the origin." The two values (A and B) extrapolated to the Z'-axis are attributed to the bulk component in the resistance and the sum of the bulk and grain components in the resistance, respectively. The complex impedance plots for the Y-stabilized zirconia connected with the composite are shown in Fig. 8. The complex impedance plot was represented by an arc which passed through the origin and spur at 250 C. At a high temperature, ca. 300 "C, two arcs were clearly observed in J.MATER. CHEM., 1994, VOL. 4 A 0 00"aa O-I I I A €3 A Z'(arb. units) Fig. 7 Complex impedance plots of Y-stabilized zirconia: (a) 300 "C, (b)450 -C 1; (b) 0 I I I I A AB Z'(arb. units) Fig. 8 Complex impedance plots of the composite connected with zirconia sintered at 1200 "C: (a) 250 "C, (b)400 "C which the arc observed at high frequency passed the origin. At higher frequency two intercepts (A and B) to the Z-axis are expected. The first intercept, A, corresponds to the bulk component to the resistance and the second, B, to the sum of the bulk and grain components to the resistance. At higher temperature, an arc and a spur were present and two intercepts were expected. The lower value of Z' corresponds to the bulk component and the other to the sum of the bulk and grain components.From these observations, the resistivity for each component was estimated and the temperature dependence of the resistivity is summarized in Fig. 9. The resistivity param- eters, E and RT, of the composite connected to Y-stabilized zirconia are comparable to those of the Y-stabilized zirconia alone. In addition, the activation energy of the bulk compo- nent was slightly lower than that of the grain component. It is concluded that the electrical properties are mainly con- trolled by the Y-stabilized zirconia layer, i.e. the conductivity of the composite layer is higher than that of the Y-stabilized zirconia layer. C0,-sensing Characteristics For the cell, expressed as: 02,Pt llNa ionic electrolyte INa2C03 IIPt, CO,, 0, 1.o 1.5 2.0 1O~WT Fig.9 Temperature dependence of the resistivity: Rbulk (a) and Rgrain (b) of the composite connected with Y-stabilized zirconia which was sintered at 1200 "C, and $?bulk (c) and Rgrain(d) of the Y-stabilized zirconia it is assumed that the chemical potential at the mode is controlled by the reaction: Na2CO,-,2Na+CO,+~O2 (9 whereas that at the cathode is represented by 2Na++02+Na20 (ii) The overall reaction is predicted to be Na,C0,-+Na20 +C02 (iii) When the oxygen concentration in the anode is kept the same as that in the cathode, the emf of the cell is expected to be = -(AGNazO +AGCO, -AGNa2C03)/2F -(RT/2F) In pCOz (2) where AGi is the standard Gibbs energy of formation and Pcoz the concentration of CO, in air.If the activity of Na20 remains constant, the emf gives the concentration of CO,. Fig. 10 shows the sensing characteristics at 470 "C for a sensor having the structure: O,, PtlZrO, llNa ionic electrolyte INa,C03 )IPt,C02,0,(iv) in which the composite connected to Y-stabilized xirconia calcined at 1200 "C was used for the electrolyte body. In this .. _.. . i -Ap.-. .-:-: , . lo , . . > 400-.. E .. 100== 7~ . _. ,-..-_.. .-...... . . ....s .... .. .... . ... ...... 1000 i, . -.300--,._... . ..... . 150 min_ ....._ -. _ . ._ 200 10000 1 Fig. 10 C0,-sensing behaviour of a sensor with the structure6 (refer- ence) PtJZrO, // Na ionic electrolyte INa2C0, // Pt (sensing electrode).CO, concentrations in ppm are denoted in the figure Fig. 11 CO, concentration dependence of the emf of the sensor with the structure given in the caption to Fig. 10. (a) Dry test gas and air in the reference, (b) dry test gas and 1OOOppm C02 gas in the reference, (c) wet test gas (dew point 30 "C) and air in the reference case, air (50ml min-') was introduced on the reference electrode side. On switching from 1000 ppm C0,-air flow to 100 ppm C0,-air flow, the emf increased rapidly and a steady- state value was observed. The rise and recovery times were very fast: the 90% response time was <2 min. When the CO, concentration was changed from 10000 ppm to 100ppm, from 100ppm to 10ppm and from 10ppm to air, the response time became longer.It seems that the concentration in the measuring cell could not be obtained soon after changing from a higher to a lower concentration. The response time was reduced by increasing the flow rate of the test gases and by increasing the number of measuring cycles. The con-centration dependence of the emf is shown in Fig. 11. A good linear relationship was confirmed and expressed by the relation, E/mV =514-72.5 log [Cco, (ppm)] in the range 10-10 000 ppm CO,. The sensitivity, 72.5 mV, is in reasonable agreement with the theoretical value, 74mV, based on eqn. (iii). In addition, when the air with 1000ppm CO, was introduced on the reference electrode side, the emf was expressed as: E/mV =5 15 -72.0 log [Cco2]in the same range.It is clear that the sensing characteristics are barely influenced by the C02 content on the reference electrode side. It is concluded that the activity of Na20 in the reference side/the interface of composite-Y-stabilized zirconia remains constant and is uninfluenced by changes in the concentration of C02 when the electrolyte is completely covered with Y-stabilized zirconia. The results suggest the possibility of fabricating a CO, gas sensor for use in ambient air without a control on J. MATER. CHEM., 1994, VOL. 4 the reference side. For a sensor expressed as: O,, Pt llNa ionic electrolyte)Na,CO, IIPt, CO,, 0, (v) in which the composite sintered at 1050 "C was used as the sodium-ion electrolyte, the emf in dry air is expressed as E/mV= 551 -74.5 log[Cco2] in the range 100-10000 ppm CO, when dry air is introduced to the reference electrode side.In this case, the sensitivity, 74.5 mV, is reasonably close to the theoretical value based on the two-electron electro- chemical reaction. When a test gas containing 10ppm CO, was used, the emf was lower than the extrapolated value estimated from the relationship confirmed at higher concen- tration ranges. Ambient air contains some water molecules, so the influence of humidity in the test gas on the sensing char- acteristics was examined. The result is shown in Fig. 11. When humid test air (30 "C dew point) was passed, the emf was lower than that measured for the dry test air, and in the range 100-10000ppm CO,, the emf is expressed as E/mV =496 -70.0 log[CCo2].Furthermore, the emf for [CO,] <100 ppm was lower than the value expected from the relation obtained in the higher range, the emf decrement increasing with decreasing CO, concentration. These dec- rements are ascribed to the formation of sodium oxides, such as Na,O and Na202, in the Na2C0, layer and/or the Pt electrode/body interface, as reported previ~usly.~~~," For the examined cell/sensor, the sensing characteristics in dry air remained almost constant even after exposure to humid air for 12 h or more. Furthermore, it is confirmed that the CO, concentration even in humid air can be measured without any distinct drift and/or loss of sensitivity (no distinct change in sensitivity was observed when sensing was carried out in humid air for 1 day or more). References 1 C. H. Kim, B. Qiu and E. Banks, J. Electrochem. SOL'., 1985, 132,1340. 2 E. Banks and C.H. Kim, J. Electrochem. Soc., 1985,132,2617. 3 K. Jackowska and A. R. West, J. Mater. Sci.,1983,18,2380. 4 D. Bahadur, Phys. Status Solidi (a), 1986,98, K23. 5 Y. Sadaoka, M. Matsuguchi and Y. Sakai, J. Mater. Sci., 1989, 24, 1299. 6 Y. Sadaoka, Y. Sakai and T. Manabe, J. Mater. Chem., 1992, 2,945. 7 Y. Sadaoka, Y. Sakai, M. Matsumoto and T. Manabe, J. Mater. Sci., 1992,28, 5783. 8 C.C.Hunter and M. D. Ingram, Solid State Ionics, 1984,14,31. 9 R. M. Hakim and D. R. Uhlmann, Phys. Chm. Glasses, 1971, 12, 132. 10 N. Matsui, Dennki Kagaku, 199 1,59,79 1. 11 J. Liu and W. Weppner, Eur. J. Solid State Inorg. Chem., 1991, 28, 1151. Paper 3/07424B; Received 17th December, 1993