Ceramics International 43 (2017) 83298333

Contents lists available at ScienceDirect

Ceramics International
journal homepage: www.elsevier.com/locate/ceramint

Glass-ceramic joining material for sodium-based battery MARK

a, b a a a a
F. Smeacetto , M. Radaelli , M. Salvo , D. Di Modugno , A.G. Sabato , V. Casalegno ,
M. Brogliab, M. Ferrarisa
a
Department of Applied Science and Technology, Politecnico di Torino, Corso Duca degli Abruzzi 24, 10129 Torino, Italy
b
RSE spa Power Generation Technologies and Materials Department, Via Rubattino 54, 20134 Milano, Italy

A R T I C L E I N F O A BS T RAC T

Keywords: This study focuses on the design, characterization and testing of a new glass-ceramic sealant for Na-based
Joining batteries. The thermo-mechanical properties as well as the sintering behaviour of the new sealant have been
Glass ceramics characterized by thermal analyses and x-rays diraction studies. The exothermic peak in the dierential thermal
Batteries analysis plots has been assigned to the crystallization of a calcium sodium aluminum silicate (anorthite sodian),
Glass
as conrmed by the X-ray diraction data. The activation energy for the crystallization was found to be
Thermal properties
Al2O3
482.3 kJ/mol. The compatibility at the interface between the glass-ceramic sealant and beta- and alpha-alumina
is examined before and after corrosion test with molten Na. Tests of this novel silica-based glass-ceramic sealant
with Na at 300 C for 250 h in a real battery operating conditions showed no reactions with Na or evidence of
corrosion, as demonstrated by SEM post mortem examinations after cell disassembly.

1. Introduction contact. In the view of these specic properties, glass-ceramics, which

Sodium-beta-alumina batteries are receiving remarkable considera-
tion as attractive devices for utility-scale energy-storage applications
[1]. The core assembly of these batteries is composed of molten
sodium/sulphur or sodium/metal chloride as electrodes and beta-
Al2O3 ceramic as the electrolyte. These batteries operate at a tempera-
ture around 300 C, exhibit high power and energy densities, tempera-
ture stability, and low cost. A reliable sealant between the beta-alumina
and the alpha-alumina is therefore necessary and required for both
safety and performance reasons [2]. Crucial properties of an eective
sealant include thermo-mechanical compatibility with the alumina,
chemical stability with Na at 300 C, excellent tightness, and reliable
joining performance during operation. Glass is typically used to seal the
electrolyte alumina-based ceramic and the insulating ring in the
sodium-beta-alumina battery; the glass sealing technology is therefore
considered as one of the most signicant in order to ensure the safety
and a good performance of the device [3]. The choice and design of
appropriate compositions of sealing glass is specic and very challen-
ging, since particular properties in the hostile environment of these
devices include the corrosion resistance against molten sodium at T
around 300 C [4]. Furthermore, the sealing glass must have an
adequate coecient of thermal expansion (CTE) to minimize the
thermal stresses and appropriate characteristic temperatures, thus
exhibiting sealant strong bonding with the materials to which is in
can be shaped by a controlled devitrication of the parent glass, are
likely to be the designated materials due to their versatility in terms of
compositions, crystalline phases and dierent amounts of residual
glassy phase [57].
Recently, few works and patents about new types of sealants, such
as glass-ceramics, Bi-doped borosilicate glass have been reported [8
11]; the partial substitution of Bi2O3 for SiO2 in a borosilicate glass was
found to improve the thermal expansion match between the glass and
the alumina substrates, increasing the glass transition temperature
(Tg), the softening temperature (Ts) and the viscosity of the doped
glass [12]..
S. Kim et al. [13] proposed a novel CaOAl2O3 joining material as
an alternative adhesive to replace the traditional glass sealants; the
adhesive has been tested for 1000 h at 180 C in a molten Na
environment. Glass-ceramic materials are typically produced through
the sintering of the parent glass powders. Crystallization is usually
benecial for a seal glass because the obtained glass-ceramic material is
typically stronger than the parent glass [14]. Sintering and crystal-
lization processes occur during the thermal treatment with a substan-
tial eect on the nal densication of glass-ceramics materials, which is
obtained through viscous ow at temperatures higher than the glass
softening point [15]. The accurate sequence of events in the sintering
process as well as the crystallization mechanism and kinetics are
critical issues which enable the formation of a dense glass-ceramic

Corresponding author.
E-mail address: federico.smeacetto@polito.it (F. Smeacetto).

http://dx.doi.org/10.1016/j.ceramint.2017.03.170
Received 16 December 2016; Received in revised form 6 March 2017; Accepted 27 March 2017
Available online 29 March 2017
0272-8842/ 2017 Elsevier Ltd and Techna Group S.r.l. All rights reserved.
F. Smeacetto et al.
Fig. 1. (a) superimposition of DTA analysis of glass powders (particle size < 38 m) heated at 20 C/min, with the sintering curve as obtained by the HSM. (b) DTA plots at dierent
heating rates (for glass powders of particle size < 38 m).
material and are relevant to the development of a heat-treatment
procedure for the production of a new glass-ceramic seal from powders
by a sintering-crystallization method [16]. It is possible to study the
crystallization kinetics and mechanisms with both isothermal or non-
isothermal methods [17]. In this study the non-isothermal method
have been used by carrying out dierential thermal analyses on glass
powders at dierent heating rates.
2. Experimental
A novel glass composition was formulated with the support of the
SciGlass database (Science Serve GmbH, Sciglass 6.6 software, 2006,
Germany), fullling criteria of CTE between 6 and 7106 and a glass
transition temperature (Tg) around 600 C. Besides the thermomecha-
nical compatibility with alumina, the choice of the glass-based sealant
composition was also based on the thermodynamic stability of the
oxides with respect to the sodium, i. e. Ellingham diagrams [18]. The
designed glass composition was found to be: 51 wt% SiO2, 7 wt% Na2O,
6 CaO, 22 wt% Al2O3, 2 wt% Li2O and 12 wt% B2O3, with a calculated
theoretical density of 2.44 g cm3. The sealant was produced as a glass,
by melting the oxides and carbonates raw materials in the right
amounts and by heating at 16501700 C for 1 h in a platinum
crucible; the melt was casted on a metal plate and the glass was
subsequently ball milled and sieved into particle size < 38 m. The
characteristic temperatures were measured by DTA (Netzsch, Eos,
Germany); three samples were used to determine the glass transition
temperature; Tg was detected at the onset, obtained by the intersection
of the two tangents at the start of the endotherm; the shrinkage (%)
behaviour vs temperature were obtained by a heating stage microscopy
(Expert System solutions, Italy). Additional DTA analyses were carried
at dierent heating rates (5/10/20/40 C/min) out in order to study
the crystallization kinetics of the glass-ceramic. Coecients of thermal
expansion (CTE) were measured on both glass and glass-ceramic
samples (554 mm3) with a heating rate of 5 C/min; glass-ceramic
samples were prepared by uniaxially pressing a mixture of glass
powders ( < 38 m) and isopropyl alcohol; the obtained pellet was
subsequently sintered at 10001100 C, for 1 h in an oven in static air
atmosphere. Three samples were used for determining the declared
CTE value. The joining procedure was carried out by depositing at RT
the glass-based slurry (a combination of glass powder+ethanol, solid
content 45%) on the alpha-alumina substrate and subsequently by
Table 1
Thermal, thermomechanical properties and kinetic parameters of the glass-based joining material.
Ec (kJ/mol) EcK (kJ/mol) n m mEc (kJ/mol)
482.3 244.5 1.9 1 482.3
Ceramics International 43 (2017) 83298333
placing the beta-alumina disc on top, in order to form a sandwich-like
structure. The joined sample was heat treated at 1000 or 1100 C in
static air for 1 h; a heating rate of 10 C/min and a cooling rate of 5 C/
min were used in the sintering runs. Initial joining tests were, useful for
preliminary tests with molten Na, were carried out by using alpha-
alumina substrates. Alumina/glass-ceramic/alumina joined samples
have been examined by scanning electron microscopy (SEM), in order
to evaluate dierences in the sintering behaviour at 1000 and 1100 C.
The apparent density of the glass-ceramic has been measured by
Archimedes method after the sintering at 1000 C and 1100 C
respectively.
The crystalline phases in the glass-ceramic were identied from X-
ray diraction data (PanAlytical XPert Pro PW3040/60, Philips)
analysed with the aid of the X-Pert HighScore software.
The initial test was performed on samples prepared by joining two
tiles of alfa-alumina. These samples were placed in contact with
metallic sodium, put in a furnace and heated 1 C/min at 300 C in
N2 atmosphere, for about 10 h.
The assembled monocell, containing the beta alumina disc joined
by the glass-ceramic to the alpha alumina ring was tested up to 250 h;
the monocell was lled with the actual reagents, sodium on the anodic
side and nickel chloride in tetrachloroaluminate on the cathodic side; it
was heated at 320 C in N2 atmosphere and 6 thermal cycles were
performed from 250 C to 320 C; so doing the joining area has been
exposed to direct contact with the reagents. The cell was disassembled
after the tests with molten Na and post mortem examinations were
carried out on the glass-ceramic joined area by using scanning electron
microscopy (SEM), energy dispersive X-ray spectroscopy (EDS)
(Merlin microscope of ZEISS).
3. Results and discussion
The superimposition of DTA analysis of glass powders (particle size
< 38 m) heated at 20 C/min, with the sintering curve as obtained by
the HSM, is reported in Fig. 1a. At temperatures above 1000 C the
glass system showed an acceptable densication behaviour as deduced
by the HSM sintering curve, while the maximum crystallization rate
(DTA plot, Fig. 1a) is detected around 1000 C. The glass transition
temperature and the crystallization peak temperature are reported in
Table 1.
Fig. 1b shows the DTA plots at dierent heating rates (for glass
nEcK (kJ/mol) Tg (C) TP (C) glass (106 C1) glass-ceramic (106 C1)
464.5 610 998 6.7 6.9
8330
F. Smeacetto et al. Ceramics International 43 (2017) 83298333

Fig. 2. Example of the evaluation of at 950 C on the DTA peak recorded at 10 C/min (a) and Ozawa plots for 950, 960, 970, 980 C chosen temperatures (b).

Fig. 3. Matusita and Sakka plot (a) and Kissinger plot (b) for glass powders < 38 m.

Fig. 4. SEM cross section of alumina-glass-ceramic-alumina joined sample obtained at 1100 C (a) and 1000 C (b) for 1 h.

powders of particle size < 38 m). The kinetics parameters (n and m),
activation energies for crystallization (Eck and Ec), thermal and
thermomechanical properties of the proposed glass-based sealant are
summarised in Table 1. n is the Avrami coecient which depends from
the mechanisms of growth, m is the dimensionality of growth of the
crystalline phase (i. e. rod-like, disc-like or sphere-like). Eck and Ec are
dened as the activation energy for the crystallization obtained by
Kissinger and Matusita and Sakka equations respectively [1921]. The
Avrami coecient can be obtained from the Ozawa equation (Eq. (1))
[22],
d{ln[ln(1 )]}
n =
d (ln ) T (1)
where, is the heating rate and is the crystallized phase volume
fraction at a chosen temperature T. was calculated as the ratio
between the fractional area under the peak (at a xed temperature) and
the overall area of the peak (Fig. 2a); it was derived from the
exothermic peak in the DTA measurement (DTA plots in Fig. 1b). So
is possible to evaluate n by DTA measurements with dierent heating
rates. The Ozawa eq. have been applied for four dierent chosen
temperatures (950, 960, 970, 980 C respectively), and the value of n
have been evaluated as the mean value between the four values
obtained for each temperature. The Ozawa plots for the chosen
temperatures are reported in Fig. 2b; for each plot the value of n
corresponds to the slope of the line. The Avrami parameter was found
to be equal to 1.9. In the case of a number of nuclei not constant during
the DTA experiment, the dimensionality of growth m can be assumed
to be equal to n1; therefore in this case has been assumed (with an
approximation) to be equal to 1 (rod-like growth) [14,23]. Knowing n
and m, is possible to apply the Matusita and Sakka equation, that

8331
F. Smeacetto et al. Ceramics International 43 (2017) 83298333

Fig. 5. XRD patterns of the glass-ceramic sealant, obtained at 1000 C, 1 h, and at
1100 C, 1 h.
allows the calculation of Ec from the shift of the crystallization peak
temperature with the heating rate:
n mEc
ln = + Const .
TP2 RTP
where TP is the crystallization peak temperature; the plot of the
Matusita and Sakka equation (glass particle size < 38 m) is reported
in Fig. 3a;Ec has been extrapolated from the slope of the line and a
value of 482.3 kJ/mol.
The Kissinger equation (Eq. (3)) has been applied too, in order to
evaluate EcK:
E in the glass-ceramic and, as a consequence, essentially reduces the risk
ln = cK + Const .
Tp2 RTp
This equation is valid only in the case of surface growth, in this case
m=n=1 and the Matusita and Sakka eq. is equal to the Kissinger eq.
The Kissinger plot used for calculation of EcK is reported in Fig. 3b.
Even if the Kissinger equation is valid only in the case of crystallization
mainly from the surface, some considerations are possible, taking into
account the nEcK and mEc products. The values for nEck and mEc are
reported in Table 1; some studies [24] recommend that the values of
nEck and mEc can be considered comparable with an error lower than
10%. These values are very similar for the system proposed here, thus
assuring the eectiveness of these results.
The activation energy is an indication of the rate of the crystal-
lization and the Avrami parameter is related to the mechanism of the
crystallization process; HSM and DTA analyses provide further details
of the order of events during a sinter-crystallization. It is important to
have knowledge of the sintering and crystallization kinetics, in order to
develop an eective heat-treatment schedule for the production of
glass-ceramics starting from glass powders, since the crystallization
can signicantly reduce the rate of densication. It is fundamental to
manage the correct order of events, thus avoiding excessive crystal-
lization that may hinder the viscous ow.
Fig. 4 shows the SEM cross section of the alumina-sealant-alumina
joined samples, as a result of the optimised heat treatment at 1100 C
(Fig. 4a) and 1000 C (Fig. 4b) for 1 h. It is clear that the porosity
decreases with increasing the sintering temperature from 1000 C to
1100 C and a high density microstructure is obtained in the glass-
ceramic joining area sintered at 1100 C for 1 h. This is due to the most
eective viscous ow, thanks to the higher temperature (low viscosity
of the residual glassy phase). This conrms the HSM measurement
showed in Fig. 1a that shows the high compliant behaviour of the
sealant. The better densication achievable by the treatment at 1100 C
has been conrmed also by the density measurements carried out by
the Archimedes method. In fact, a value of 2.16 g cm3 was measured
for the sample sintered at 1000 C for 1 h, while a density of
2.34 g cm3 was obtained for the glass-ceramic processed at 1100 C
for 1 h. Finally, the sintering temperature of 1100 C was chosen to
produce the sample submitted to molten Na exposition, in order to
promote the devitrication and the least possible porosity formation.
(2) The interface between the sealant and the alumina substrates is free
of defects and continuous. The lack of cracks at the interfaces, in the
joining area and in the alumina substrates validates the excellent
thermomechanical compatibility between the glass-ceramic and alu-
mina. The recorded CTE values were 6.7106 K1 and 6.9106 K1
(T range 200400 C), respectively for the glass and the glass-ceramic
sintered at 1100 C for 1 h. The CTE value of the glass-ceramic is
slightly lower than the alumina substrate; this ensures a compression
of cracks formation during the cooling process.
(3)
Bright rodlike crystals embedded in the dark glass matrix are
clearly evident (Fig. 4a and b) in the sealing joining area and at sealant/
alumina interfaces. This experimental evidence is in agreement with
the calculated value of m equal to 1, which suggested a crystallization
with a rod like dimensionality of crystal growth.
The XRD patterns of the glass-ceramic sealant, obtained at 1000 C,
1 h, and at 1100 C, 1 h are shown in Fig. 5. The diractograms

Fig. 6. SEM cross section of the alumina-sealant-alumina joined sample (a) and particular of the interface -alumina/sealant (b), after test at 300 C Na for 250 h.

8332
F. Smeacetto et al.
indicated an overall single calcium sodium aluminum silicate phase
that could be indexed with the Ca.66Na.34Al1.66Si2.34O8 anorthite
sodian, card number 0861650 available in the X-Pert software
database. It can be seen that both diractograms show the presence
of an amorphous halo, symptomatic of the existence of a residual glassy
phase, benecial for the achievement of a higher densication when the
system is processed at 1100 C, a temperature where the residual
glassy phase possesses a lower viscosity.
The results of the disassembled battery after the tests, i. e. SEM
cross section of the alumina-sealant- alumina joined sample are
reported in Fig. 6a; no evidence of corrosion by the molten Na were
detected on the sealant and at the sealant-alumina interfaces as
expected by the Ellingham diagrams, otherwise detachment from the
substrates would have been occurred. This behaviour was due to the
strong bonding between the alumina and the glass-ceramic sealant thus
precluding molten Na from penetrating into the beta alumina-sealant
and alpha alumina-sealant interfaces and causing degradation. A detail
at higher magnication of the interface with the -alumina is reported
in Fig. 6b; it is possible to observe that no cracks or defects are present,
thus conrming a very good interfacial bonding, eective to prevent
any corrosion phenomena.
4. Conclusions
A new silica-based glass-ceramic was designed, prepared and used
as a sealant to join alumina substrates in a Na-based battery. The
proposed sealant was produced as a glass, submitted to a joining heat
treatment that determined the nucleation and growth of an anorthite
sodian crystalline phase and a presence of residual glassy phase.
The sealant demonstrated an excellent thermo-mechanical compat-
ibility with the alumina substrates and was found to be eective in
terms of corrosion resistance to molten Na for 250 h in real battery
operating conditions.
References
[1] X. Lu, G. Xia, J.P. Lemmon, Z. Yang, Advanced materials for sodium-beta alumina
batteries: status, challenges and perspectives, J. Power Sources 195 (2010)
24312442.
[2] S. Song, Z. Wen, Q. Zhang, Y. Liu, A novel Bi-doped borosilicate glass as sealant for
sodium sulfur battery. Part 1: thermophysical characteristics and structure, J.
8333
Ceramics International 43 (2017) 83298333
Power Sources 195 (2010) 384388.
[3] Z. Wen, Y. Hu, X. Wu, J. Han, Z. Gu, Main challenges for high performance NAS
battery: materials and interfaces, Adv. Funct. Mater. 23 (2013) 10051018.
[4] K. Jung, S. Lee, G. Kim, C.S. Kim, Stress analyses for the glass joints of
contemporary sodium sulfur batteries, J. Power Sources 269 (2014) 773782.
[5] I.W. Donald, Preparation, properties and chemistry of glass- and glass-ceramic-to-
metal seals and coatings, J. Mater. Sci. 28 (1993) 28412886.
[6] A.A. Francis, R.D. Rawlings, A.R. Boccaccini, Glass-ceramics from mixtures of coal
ash and soda-lime glass by the petrurgic method, J. Mater. Sci. 21 (2002) 975980.
[7] A.K. Varshneya, Fundamentals of Inorganic Glasses, Second ed., Society of Glass
Technology, Sheeld, 2006.
[8] S. Song, Z. Wen, Y. Liu, X. Wu, J. Lin, Bi-doped borosilicate glass as sealant for
sodium sulfur battery, J. Non-Cryst. Solids 357 (2011) 30743079.
[9] W.G. Budgen, P.R. Smith, UK Patent, GB2 207 545, 1989.
[10] Y. Akihiko, T. Tomonori, M. Makoto, EP Patent, 0 729 923, 1996.
[11] D. Park, C. Park, US Patent, 4 268 3 13, 1981.
[12] S. Song, Z. Wen, Y. Liu, The eect of substitution of Bi2O3 for alkali oxides on
thermal properties, structure and wetting behavior of the borosilicate glass, Mater.
Lett. 64 (2010) 10251027.
[13] S. Kim, J.H. Joo, S.-D. Kim, S.-K. Woo, Evaluation of CaOAl2O3 adhesive
bonding properties for -Al2O3 solid electrolyte sealing for alkali metal thermal
electric converter, Ceram. Int. 39 (2013) 92239227.
[14] H. Abdoli, P. Alizadeh, K. Agersted, Fabrication and sealing performance of rare-
earth containing glassceramic seals for intermediate temperature solid oxide fuel
cell applications, Ceram. Int. 40 (2014) 75457554.
[15] A.G. Sabato, G. Cempura, D. Montinaro, A. Chrysanthou, M. Salvo, E. Bernardo,
M. Secco, F. Smeacetto, Glass-ceramic sealant for solid oxide fuel cells application:
characterization and performance in dual atmosphere, J. Power Sources 328 (2016)
262270.
[16] E. Bernardo, G. Scarinici, E. Edme, U. Michon, N. Planty, Fast sintered gehlenite
glass-ceramics from plasma-vitried municipal solid waste incinerator y ashes, J.
Am. Ceram. Soc. 92 (2009) 528530.
[17] M. elikbilek Ersundu, A.E. Ersundu, M.T. Soltani, M. Baazouzi, Crystallization
kinetics of new heavy metal oxide glasses within the Sb2O3Na2O-WO3-PbO system,
Ceram. Int. 43 (2017) 491497.
[18] H.J.T. Ellingham, Reducibility of oxides and sulphides in metallurgical processes,
J. Soc. Chem. Ind. -Trans. Commun. 63 (5) (1944) 125133.
[19] H.E. Kissinger, Variation of peak temperature with heating rate in dierential
thermal analysis, J. Res. Natl. Bur. Stand. 57 (1956) 217221.
[20] K. Matusita, S. Sakka, Kinetic study on non-isothermal crystallization of glass by
thermal analysis, Bull. Inst. Chem. Res. 59 (1981) 159171.
[21] A.A. Francis, R.D. Rawlings, R. Sweeney, A.R. Boccaccini, Crystallization kinetic of
glass particles prepared from a mixture of coal ash and soda-lime cullet glass, J.
Non-Cryst. Solids 333 (2004) 187193.
[22] T. Ozawa, Kinetics of non-isothermal crystallization, Polymers 12 (1971) 150.
[23] L. Lilensten, Q. Fu, B.R. Wheaton, A.J. Credle, R.L. Stewart, J.T. Kohli, Kinetic
study on lithium-aluminosilicate (LAS) glass-ceramics containing MgO and ZnO,
Ceram. Int. 40 (2014) 1165711661.
[24] M. Erol, S. Kucukbayrak, A. Ersoy-Mericboyu, The application of dierential
thermal analysis to the study of isothermal and non-isothermal crystallization
kinetics of coal y ash based glasses, J. Non-Cryst. Solids 355 (2009) 569576.