Isothermal Crystallization of a Solid Oxide Fuel Cell Sealing Glass by Differential Thermal Analysis Teng Zhang, Richard K. Brow, w Signo T. Reis, and Chandra S. Ray Department of Materials Science & Engineering, Missouri University of Science and Technology, Rolla, Missouri 65409 The crystallization kinetics of a solid oxide fuel cell sealing glass were studied using a new isothermal differential thermal analysis (DTA) method. The weight fraction of glass crystallized after an isothermal heat treatment was determined from the DTA crys- tallization peak area and the crystallization kinetic parameters were determined using the classical Johnson–Mehl–Avrami equation. The glass, an alkaline earth–zinc–silicate composi- tion, crystallized in the temperature range between 7401 and 9501C. The activation energy for crystallization varied with glass particle size and decreased from 570725 to 457730 kJ/ mol as the average particle size decreased from 425–500 to B10 lm. The activation energy for crystallization, E, increased from 520720 to B600720 kJ/mol when glass particles (45–53 lm) were mechanically mixed with 10 vol% of micrometer-sized Ni or YSZ powders. This increase in E reflects the effect of a second phase in composite seal systems, but is independent of the chemical nature of the additives. The measured values of the Avrami exponent (n) indicate that surface crystallization is the dominant crystallization mechanism for this glass, particularly for small particle sizes, e.g. n 5 0.970.1 for B10 lm. I. Introduction P LANAR designs of solid oxide fuel cells (SOFC) require an insulating, high-temperature hermetic seal, often between a ceramic component of the cell (e.g., the Y-stabilized zirconia (YSZ) electrolyte) and the ferritic steel interconnect, to separate the fuel gas and oxygen supplies. There are many requirements for the sealing design that need to be fulfilled. For example, the glasses must have a thermal expansion match to the fuel cell components, must be electrically insulating and must be the- rmochemically stable under the operational conditions of the stack. The seal also should be chemically stable with other cell components, should be stable under both the high-temperature oxidizing and reducing operational conditions, and should be created at a low enough temperatures to avoid damaging other cell components (under 9001C for some materials). In addition, the seal material should flow enough to seal the edge during the sealing process and still retain the mechanical strength to support the substrates. Finally, the sealing system should be able to withstand thermal cycling between the operational temperature and room temperature. There are a number of possible joining and sealing tech- niques, including brazing, and sealing with glasses and glass–ceramics. 1 The glass–ceramic materials offer the best opportunity to satisfy the seal requirements because they can provide the requisite thermal expansion characteristics and the viscosity of the residual glass can be modified by controlling the crystallization process. 2 In addition, glass–ceramic materials often possess better thermochemical stability and greater mechanical strength than typical glasses. It is well known that the crystallization process plays an important role in determining the properties and applications of glass–ceramic sealants. For example, an installation process for the Siemens-SOFC stack required that the sealing glass be partially viscous at 9501C for 2–3 h to allow small displace- ments of the single stack elements after joining at 10001C. This can be achieved by using a slowly crystallizing glass. 3 Lara et al. 4 advised that the sintering stage should be completed before significant crystallization occurs to get a fully dense material suitable for an SOFC seal. Uncontrolled crystallization during the initial sintering process can lead to the formation of a porous sealing layer that can adversely affect the SOFC operation. The viscosity of glass–ceramic materials used for SOFC seals is affected by the crystallization process of glasses. Eichler found that the addition of MgO to barium aluminosilicate glasses accelerates the crystallization process of specific phases and therefore increases the viscosity of the residual glass. 3 Similarly, Sakaki et al. 5 observed an increase in seal viscosity at about 9501C because of bulk crystallization of wollastonite (CaSiO 3 ) in the CaO–Al 2 O 3 –SiO 2 glass system. The coefficient of thermal expansion (CTE of a glass–ceramic depends on the formation of specific crystalline phases from the glass. Lahl et al. 6 reported that the formation of detrimental crystalline phases like Mg 2 Al 4 Si 5 O 18 , with a relatively low CTE (2  10 6 K 1 ), in MgO-containing glasses can be suppressed by reducing the Al 2 O 3 content in the base glass, as well as using appropriate nucleation agents like Ni and Cr 2 O 3 which yield high activation energies of crystallization and crystal growth. The formation of BaZrO 3 (CTE of 7.9  10 6 K 1 ), in compos- ites of BaO–CaO–MgO–B 2 O 3 –Al 2 O 3 glass and YSZ fillers, reduced the CTE from 9.4  10 6 to 7.6  10 6 and 8.0  10 6 K 1 with nanoscale (B27 nm) and micrometer-size (o53 mm) YSZ additives, respectively. 7 The kinetics for the overall glass crystallization is typically de- scribed by the Johnson–Mehl–Avrami (JMA) 8 model, which is applicable only under isothermal heat treatment conditions. There also exist several non-isothermal models for crystallization 9–13 of which the Kissinger model 9 is most widely used. Nonisothermal approaches are often favored because of their operational simplic- ity and convenience, compared with isothermal models, in addition to being less time consuming. A modified form of the Kissinger model, introduced by Matusita and Sakka, 14 distinguished crys- tallization processes that occur on a fixed number of nuclei from those where nucleation and crystal growth occur simultaneously. The popular nonisothermal models are based on the JMA model for isothermal transformation kinetics. This raises serious questions about the conclusions drawn from typical nonisother- mal DTA experiments. It has been demonstrated that noniso- thermal experiments, like those based on the Kissinger equation, may be appropriate only over a limited temperature range when crystal growth rate increases linearly with increasing tempera- ture, and yield reasonably accurate results provided that the ex- periments are performed strictly within this temperature range. 15 Recently, a new DTA method was developed that uses con- venient, nonisothermal DTA experiments to analyze the effects J. Stevenson—contributing editor This work was supported by the Department of Energy/SECA (Project NT42221). w Author to whom correspondence should be addressed. e-mail: brow@mst.edu Manuscript No. 24447. Received March 21, 2008; approved July 24, 2008. J ournal J. Am. Ceram. Soc., 91 [10] 3235–3239 (2008) DOI: 10.1111/j.1551-2916.2008.02661.x r 2008 The American Ceramic Society 3235