P0WBm TrdIilNml~

ELSEVIER Powder Technology 88 (1996) 173-178

Effect of slurry properties on anode cermets for solid oxide fuel cells

S. Sddhar, U.B. Pal Massachusetts Institute of Teehnotogy, 77 Massachusetts Avenue, Cambridge, MA 02! 39 USA

Received 21 September 1995; revised I Febvm~ 1996

Al~tract

The influence of the initial slurry preperties on the resulting Ni-YSZ ( y'~rla-stshiiized zirconia) cermet structure that constitete~ the solid oxide fuel cell anode has been examined. Rheological study of the slutr~ '-::s been conducted as a function of powder concentration. The powder consisted of Ni (80 wt.%) and YSZ (20 wt,%) particles, and the s!urry was prepared by mixing the powders in polyvinyl alcohol (PVA) solution, This suspension was shown to Ig pseudoplastic without a yield point and a model based on the Ostwald- de Waele power law has been developed. Porous structures resulting frvrn slurry coating, sintering and infiltration were examined utilizing microprobe. scanning electron microscopy (gEM) and X-ray techniques. The resulting structure was shown to be inadequate when the slurry suspensions contained less than 12 or more than 13 vol.% of the powders.

Keywords: SlunT; Pseudopiesfic; Settling: Cezw.et; Solid oxide fuel cell; Electrode

1. Introduction

A porous cermet of Ni and YSZ (yttria-stabilizedzircouia) constitutes the anode in state-of-the-art solid oxide fuel cells (SOFC). The role of ~he structure is to provide current col- lectors (Ni particles) and electrochemical charge transfer sites (three phase interfaces between Ni. YSZ and pore). Furthermore, the anode often assumes the role of a catalyst for various chemical reactions. To maintain the polarization losses at acceptable levels, it is imperative that the Ni and YSZ particles form continuous paths that allow ionic and electronic migration from the electrolyte/cermet interface through the entire cermet [ ! ]. In addition, there has to be enough contact points between YSZ and Ni particles to Wo- vide charge-transfer sites.

Processing techniques for depositing these electrodes include screen printing or tape casting a layer of NiO and YSZ over the electrolyte followed by reduction of the NiO to Ni [2]. These processes are sensitive to the particle size of the NiO because the contact area between the particles changes during the course of the reduction process. Further- more, the reduced Ni particles often have a core of NiO. As a result, these electrodes have higher sheet resistance. A Com- bination of slurry coating and electrochemical vapor depo- sition (EVD) [3] or physical vapor deposition (PVD) is often used as an alternative method to depe6it the cermet.

0032-5910/96/$15.00 O 1996 Elsevier Science S.A. All tights te~e~ed PIIS0032-5910(96)03114-2

These processes result in stable cermets, however they also involve expensive vapor deposition processes. Slurry coating is an economically favorable method that has shown to give reproducible results in terms of porv6ity and layer thickness [4,5]. The process consists of coating die electrolyte-sub- strate with the slurry and, after drying, subjecting it to a number of sintering cycles. The initial slurry and coating/ drying process has a great degree of influence on the final porosity and panicle dis~hnfiou. Hence it is imperative that the properties of the slurry are closely monitosud. The advan- tage of studying and controlling theological properties are twofold. Firstly, the optimal slerry/powder concentration and dispersion technique can be established and secondly, changes in sluny properties dining sU3rage due to evaporation of water and settling of heavy particles (Ni) can be monitmed and the col~¢ntrations alKI dispersions can be altered to obtain the required theological properties for repredocibly depositing the desired cermet stroctme.

2. Material and experimental procedure

The filamentary Ni powder used was INCO 287 with a diameter of 2.2-3.3 pan and a length of 30-50/an. "lids powder type has Igen used in earlier work [4,5] due to its excellent conducting prot_~.~es and good sintering capabili-

174 s. Sridhar, U.K Pal/Powder Technology88(1996) 173-178

tics for achieving controlled porosity. YSZ was obtained from TOSOH-Ceramic Division and consisted of 13.47 wt.% Y~O3 doped ZrO 2. Particle analysis showed the particle size distri- bution to be between 9 and 0.2 pin with over 90% of the particles being smaller than 3 /tm. The polyvinyl alcohol (PVA) solution was prepared by heating deionized water containing 6 wt.% PVA powder (Dupont, EIvano175-15 ) to 80 °C with continuous stirring. The slurry solutions were prepared by dispersing known mixtures of Ni and YSZ pow- ders (Ni constituted 80 wt.% of the powders) in a known volume of PVA solution. The powder was added in four batches. Between each addition the dispersion was milled in a SWECO-Vibromill for 5 rain. After all the powder was added, the dispersion was milled for 66 min. Milling for longer times did not result in any change in the theological properties. Dispersion by manual stirring required much longer times to achieve the same rheological properties. No grinding media was used since this would cause the soft filamentary Ni particles to break.

Rheological characterization of the slurry at various pow- der to PVA solution ratios were conducted at room temper- ature with a Brookfield viscometer, model DV-II +. A small sample adapter was used to hold 8 ml of the slurry to be tested. For the viscosity range encountered, spindle #21 was most applicable. Due to the fast settling rate of the Ni parti- cles, the viscosity had to he measured immediately (within 1 rain) after the milling process. Measurements were made while ramping from the lowest to highest shear rate and then ramping back to the lowest. At high shear rates, measure- ments were made after 10 spindle rotations. However, due to the t'~st settling rate this could not he done at the low shear rates ( corresponding to low rotation speeds ), hence the meas- urements were taken after 40 s.

After milling, the slurry was immediately coated on one side of a YSZ-electrolyte substrate. The coating was then passed through a blade, adjusted at a constant height to achieve an even coating after which it was left to dry for 24 h. The coated substrate was then sintered at 1100 °C for 17 h under an atmosphere of forming gas (N, with 5% Hz) with 5% H,_O followed by cooling under an atmosphere of pure forming gas. To increase the YSZ content especially at the electrolyte/cermet interface and obtain long time adher- ence, the substrate was then placed inside a vacuum chamber and infiltrated with a solution containing 30 vol.% YSZ. and 70 vol.% of a deionized water and 4N HNO3 mixture. Finally the coating was sintered at 1200 °C for 7 h in forming gas. The final cermet structure after this operation usually con- tained 60 vol.% Ni which is above the limit (30 vol.%) required according to the percolation theory for the cermet electronic conductivity to be close to that of Ni [ 6].

The YSZ in the initial powder and more so the YSZ in the infiltration solution contributes to the stability of the cermet and its adherence to the electrolyte. Due to incompatible thermal expansion, Ni will not by itself adhere to YSZ sub- strates at high operating temperatures of the SOFC (800- 1000°C).

3. Results and discussion

3.1. Rheologicalproperties

Viscosity as defined by Skelland [7] is non-Nowtonian, that is, dependent on shear rate for most suspensions and especially so for multi-phase slurry solutions with tendencies for settling.

,/d, /

The deviat ion f rom Newtonian behavior, fo r example, the dependence o f ¢) on shear rate, w i l l characterize the solution as being pseudoplastic, dilatant, Bingham, etc. Moreover, time dependency and yield stress phenomenon may be present.

Figs. 1 and 2 show the measured rheological properties of Ni-YSZ powders suspended in a PVA solution. The powder mixture consisted of 80 wt.% Ni and 20 wt.% YSZ. Volume fractions of powder, Xp, between 0.091 and 0.132 were studied.

~o. £ •

t: o i t ) i i

o 20 40 ~o 80 lOO shear Rate (lhs)

Fig. I. Shear stress vs. shear rate for different volume fractions (Xp) of Ni- YSZ po',,, der suspended in a PVA solution.

8000 ] ~ = 0 . 1 0 t ] I'--A'--Xp=I3.112 I

7ooo I "N-xp< ' l= t I

= o 0 ' I ' - * - x P = ° ' t = I

lOOO a ~

tO 20 30 40 50 ah~w ~ (11/=)

Fig. 2. Viscosity vs. shear rate for different volume fractions of power (Xp). shown in Fig. I.

7OOO

A S ~

i =

Ioi~.

o

S, Sridhnr. U.R Pal/Powder Technology 88 (1~6) 173-178

t e r ~

i , t i 10 20 ~0 40

ffaew I~lm 0/*)

Fig. 3. Thixotropic behavior for slurcy solution with Xp = 0.112, Arrows indicate the direction that the shear rate was ramped.

Some extremely low and high viscosity values have been omitted in Fig. 2 in order to maintain a better scale on the graph. It is readily seen in Fig. I that no yield stress phenom- enon is present for any of the powder concentrations studied. Moreover, since the slopes decrease as the shear t .ieincreases the fluid can be characterized as pseudoplastic (shear thinning).

This behavior can also be deduced from Fig, 2 where the viscosity is seen to decrease with shear rate.

Most of the samples displayed some degree of hysteresis when the shear rate was ramped back. The nature of the hysteresis is shown in Fig. 3. The type of hysteresis was consistent for all the samples that showed time dependency.

Of the many existing models for l~eudoplastie fluids [7] the Ostwald-de Wanle model, described by Eq. (2). and also called the power law model, is the most commonly obeyed and seems to give a reasonably good prediction of the thee- logical properties of the Ni-YSZ powder slurry.

~dy! (2)

where "r is the shear stress, K is the consistency factor and n is the flow behavior index which changes from unity to zero as pseudoplastieity increases.

From Eqs. ( I ) and (2) the viscosity can be written as:

de m ~ / = / ~ ) w h e r e t o = n - 1 (3)

By plotting the log of viscosity or shear stress versus shear rate the validity of the model can be tested and n. m and K can be evaluated.

Table I shows the values obtained through least square fitting of the data in Figs. 1 and 2. Note that m should equal (n - 1 ) and g ( in eP) shoeld equal/¢ × 100. Table I indicates

175

Table I Paralnete~ for the Oslwald-de Waale model evaluated flora experir~ntal

.400 dala

! 350 ~ Xp n g m K r,/C n-I (dynes/cm 2 ) (cP) (m).tcP)

R l 0,132 0.52 93.8 - 10.53 9286 0.961 -0AS

250~ ~ 0,128 0.57 84.0 -0.49 8800 0.973 -0.43 0.121 0.52 52.9 -0.47 5380 0.994 -0.48

200 0.112 0.54 47.9 -0.46 4833 0.993 -0.46 0.10l 0.68 11.3 -0.31 lit6 0.997 -0.32

150 ] 0.091 0.74 6.51 -0.26 645 0,994 -0.26 lm t4

that n and m are decreasing functions of Xp whereas K is increasing.

As a first approximation, the data has been fitted as a linear function with respect to the volume fraction of I~rlficles, Xp. Eq. (3) can thus be written as a function of Xp, namely:

"~ = ( - 20.4 X 103) + (22.3 X 104 X Xp) ( ~ ) ~'s-~'~xxp

(4 )

3.2. Characterizationoffmaleermetstrucmre

Several slurries with powder volume fraction between 0.09 and 0.135 were deposited and sintered according to the tech- nique described in the previous section. The blade for the initial coating process was adjusted so that the final film was 60-70/~m thick. An environmental SEM was used to char- acterize the deposited films using energy dispersive X-ray spectroscopy. Several areas of 300×250 p,m 2 were exam- ined. The results showed that. at lower powder concentra- tions, cermet free areas leave the electrolyte substrate directly visible. Increasing the water/PVA content in the slurry (for Xp<O.12) doss not seem to change the cermet itselfbut rather increases the cermet free regions on the electrolyte substrate. Fig. 4 and Fig. 5 show characteristic areas from two samples with X v of O.(Y) and O. 13, respectively. It is desirable to cover the entire electrolyte surface with the eemlet` and hence frotn this point of view it is preferable that the slun 3' has a high powder content, Above Xp=0.12 most of the eleetrolyte seems to be covered. Microprobe imaging of cress-sectional areas of the cermet showed that the fraction of voidel area was around 17% and did not vary much for slurries with Xp between 0.09 and 0.135.

Figs. 6-8 show X-ray maps of Hi and Zr in a cross-sectional area of the cermet for Xp = 0.09, O. I 12 and O. 132 respeotively. It can be seen that at X v = O. 132 the Ni segregates and agglom- erates near the electrolyte and also undergans extensive sin- tering. Due to their high density, the Ni particles settle to the bottom of the cermet` thereby increasing the Ni concentration near the electrolyte region. Furtbermore as the water evapo- rates, it brings the lighter YSZ particles to the surface while leaving the heavier Ni particles underneath. The infiltration proeeas after the fast siotering ensures that the depletion of

176 S. Sridhar. U.R Pal / Powder Technology 88 (1996) 173-178

~i~!iii,i~ii~i!~ii/~il;~ii~ ¸ / ~ ! i ~ / i I

ii!i!i~!i!iiii~!~! i~!i~ %ii ̧ i~ ~ : i i~ i ! i~i ~i~ii!@i~ii~!!!~iiill i !iiii:~ ~, i~ ¸¸ , !!~iii:iiiii ~i!ili%~i!!i !! iiili` / ¸ i !i~ i ~i~!!!/il

Fig. 4. (a) SEM image of cermet processed from a slurry with Xp= 0.09; (b) X-ray map of Ni (left-hand side) and Zr (right-hand side).

, =

~ii~ ~

i Fig, 5, (a) SEM image of cm'n~t processed from a sluny wilh Xp~ 0.13; (b) X-ray map of Ni ( ieft-h~d side) and Z¢ ( right -hired side).

YSZ is compensated with the YSZ rich infiltration solution. However, if porosity near the electrolyte substrate is low after the first sintering dee to a high Ni powder concentration, the

Fig, 6. Cross-sectional area of a cermet processed from a slurry with Xp ~ 0.09. (a) Xoray map of Ni; (b) X-my nmp of Zr.

YSZ solution that is intended for infiltrati:~n does not reach the cermet/electrolyte interface. The second sintering causes further Ni sintering resulting in an almost compact Hi layer at the interface (see Fig. 8). Since the thermal expansion coefficients of Hi ( = 14 om/cm K) and YSZ ( = 9 cm/cm K) am incompatible the resulting adhesion between the cer- met and the substrate is low. Indeed some of the corrects made from slurries with Xp>0.13 pealed off from the sub- strates in the form of a thin compact film.

Fig. 9 essentially summarizes Eq, (4) and defines the region of desired powder concentration: O.12<Xp<O.13. Below 0.12, large parts of the electrolyte surface are uncov- ered whereas above 0.13 the adherence of the cermet to the electrolyte is poor due to the segregated Ni particles.

4 , C o n c l u s i o n s

The rbeological behavior cf slurries for SOFC anode cer- mets have been investigated for volume fractions of powder, Xp, bet ween 0.09 and 0.132. The powder consisted of 80 wt.% Ni and 20 wt.% YSZ. The aluny is pseudoplastic and agrees with the Ostwald-de Waele power law. Some degree of time

S. Sridhar, U.R Pal /Powdcr Technorogy ga (1996) 173-178 177

Fig. 7. Cross-sectional ar~a of a ¢~llnet processed from a s|uny with gp= 0.112, (~t) X-ray map of Ni; (b) X-ray leap of Zr.

dependency is seen but it is hard to draw further conclusions, since this may be due to the aligning of the filamentary Ni particles or simply settling of the same. Slurries containing Xp=0.09, or less, caused large areas on the substrate to be without cermet. These uncovered areas decreased in size as the powder amount was increased and above Xp =0.12 most of the substrate was covered. At Xp = 0.132 and higher con- centrations, settling and sintering of Ni and depletion of YSZ at the electrolyte substrata/cermet interface caused peer adherence.

5. List of symbols

A area (dynes/era 2) F force (dynes) K consistency factor (dynes/cm 2 or c+P) m n - I n flow behavior index r regression factor v velocity (em/s)

Fig. 8+ Crms.s=ctio~ area of a c ~ p~-e~ed ~ a ~ w~ Xp=0,132. (a) X-my map ofNi; (b) X-ray map of Zr.

"t

O.~I 0.1 0.11 0.12 6.13 0+14 0-15 0+~ Xp

X~g, 9. Slmy V i ~ t t y at diltemmt sh~m" ram as a fmmtion of ~ n mo~ing to Eq. (4) and its effmt cm cermet ~

Xp volume fraction powder y separation (era)

Greek letters

,{ viscosity (cP) shear stress (dyacs/em 2)

178 S. Sridhur, U.B. Pal/Powder Technology 88 (1996) 173-178

Acknowledgements

Financ ia l s uppo r t by the Elec t r ic P o w e r Resea r ch Inst i tute ( E P R I ) is g r a t e fu l ly a c k n o w l e d g e d .

R e f e r e n c e s

[ I ] T Kawada, N. Sakai, H. Yokokawa and M Dokiya, S~did State hmics, 40/4-1 (1990) 402.

[ 2 ] P.H. Middleton, M.E. Seierslen and B.C.H. Steele, in S.C. $inghal (ed.), Proc. 1st Int. Syrup. on Solid Oxide Fuel Cells, The Electrochemical Society, Penninglon. NL 1989, pp. 90-98.

[3 ] A.O. lsenberg and G.E. Zymboly, US Pa:¢m No. 4 582 766 (1986). [4] R.E. Jenscn, US Patent No. 4 971 830 (1990). [ 5 ] K.C. Chou, S. Yuan and U.B. Pal.in S.C. Singhal and H. Iwahara ( eds. ),

Proc. 3rd Int Syrup. on Solid Oxide Fuel Cellx, The Electrochemical Society, Pennington, N J, 1993, pp. 431--443.

[6] N.Q. Minh, J, Ceram $oe., 76 ([993) ~fi3 [ 7] AH.P. Skelland. Noa.Newtonian Flow and Heat Transfer, Wiley, New

York. 1967.