FUEL CELLS small hydrogen blower was used, and to prevent hydrogen leakage, a glandless drive was also used. A magnetically driven pump using a sealing shroud of thin nonmagnetic metal has been successfully employed for some time; it can be seen mounted underneath the battery (figure, page 302). T h e rate at which the condensate is removed from the system is controlled by switching the blower on and off at intervals. The switch is controlled by a second differential pressure meter which operates on the difference between the hydrogen in the system and the electrolyte. I n this way, the removal of water is controlled by the total volume of electrolyte which should, of course, be kept approximately constant. The condensate collects in a small vessel, from which it is released periodically by a level-sensing device
such as a capacitance probe. The main parts of this gear have been in operation and appear to work perfectly well. Until more experience is obtained, the main level gages will be retained, but eventually it should be possible to remove them. The initial heating of the battery is accomplished by electrical heaters mounted on the end plates and round the main body of the battery inside the lagging. I t is difficult to prevent some generation of hydrogen and oxygen by electrolysis in the common electrolyte ports, although insulation with p.t.f.e. helps considerably. Moreover, there is always the possibility that an electrode may leak and allow gas to get into the electrolyte system. This is taken care of by a level-sensing device, which by
means of a solenoid operated valve, will release any gas which may collect at the top of the electrolyte system. These controls may seem somewhat complicated and expensive, but they will work. With a larger battery they should be no more complex and therefore would represent only a small proportion of the cost for the whole plant. RECEIVED for review September 15, 1959 ACCEPTED December 28, 1959 Division of Gas and Fuel Chemistry, Symposium on Fuel Cells, 136th Meeting, ACS, Atlantic City, N. J., September 13-18, 1959.
F.
T.
BACON'
National Research Development Corp., London, England
Present address, Marshall's Flying School, Ltd., Cambridge, England.
High Temperature Fuel Cells E X P E R I M E N T A L WORK in these laboratories on high temperature fuel cells was initiated by a thorough investigation of the type of cells described by Davtyan (7). Because neither the electrodes nor electrolyte were found to be stable, these cells are not suitable for long term operation. Therefore, work was directed toward developing a cell that would fulfill stability requirements over long periods in the presence of reacting gases and their combustion products. The electrolyte selected was a mixture of lithium, sodium, and/or potassium carbonate, impregnated in a porous sintered disk of commercial magnesium oxide. After a presintering stage a t 1200' C., sintering was carried out at the Same temperature. Because of the low firing temperature, the disks were not very hard. They had a volume porosity of 40 to 5oy0. After impregnation, the alkali carbonate content was usually about 4oy0 by weight, which indicates that not all pores of the magnesium oxide were filled with electrolyte. The cell construction is illustrated in Figure 1. The impregnated disk (4)was covered on both sides with thin laycrs of metal powders (5) prepared by reducing the corresponding oxides in a hydrogen or carbon monoxide atmosphere. To prevent sintering during operation, the reduction temperature selected was equal to or slightly above operating ternperature. Because electrode reactions are confined to areas close to the geometrical three-phase lines bordering the metal, .electrolyte and gas phases, powder lay-
ers must be kept thin as possible and yet permit electrical contact and prevent flooding. These layers can be made less than 1 mm. thick if they are covered with silver wire gauze (O), or a t the anode with iron, nickel, or copper gauze. To prevent deformation, the gauze in turn is covered with firm pekforated stainlesssteel disks (7), 1 mm. thick. Terminal wires (9) of silver were screwed (10) into the steel disks, and the assembly was completed by gaskets of mica (8) and asbestos (3), together with steel covers (2) and pipes (11, 13) for circulating the gases. Screws (1) with appropriate mica isolating rings held the cell parts together.
Cells of this type can be run continuously for several months, operating between 550' and 700' C. on town gas, hydrogen, carbon monoxide, and natural gas. T h e best cell had a life of 6 months without an appreciable decrease in galvanic activity. However, slow deterioration was caused by a gradual loss of carbonate melt, both by direct vaporization of carbon dioxide, lithium oxide, and sodium and potassium monoxides and by chemical reactions with the gasket. The increasing loss of fused salts led to increasing gas leakage across the disks, thus lowering cell voltage. Therefore, conclusions about performance of elec-
Figure 1 . The high temperature cell. The electrolyte is a mixture of lithium, sodium and/or potassium carbonates impregnated in a sintered disk of magnesium oxide VOL. 52,
NO. 4
APRIL 1960
303
1- 2 - 3
2L'hdoy H .~ +..L. .~ l o. H z O.../ 0+.zI o l o H.-~O
t.5oooc
1 An CO2 Oole
2
9%
"
I
1Coth,COz 990 ..
/
9%
1-
C O ~ ~ Smz ~ 3w0 I ~ HzO 5 9 e Hzo C0,48% Hz0 5% C H
0
A&
Cat-
I
i
m L I V ~ oz
5
57%
co2 38%
HzO 5%
Pt lkmz n~ 10cm2
t
20
LO
60
80
tin
100
-
t
1LO
160
mA/cm2
'
Figure 2. After correction for ohmic drop, curves 1 and 2' do not show polarization. The anode i s platinum and the cathode silver
trodes and fuel gases were usually drawn within the first month of operation. To find suitable electrode metals, cells having two electrodes of the Same composition but of different areas (about 5 to-1) were studied. Terminal voltage (V,) us. current density (i), characteristics which indicated polarization differences, could be obtained by passing fuel gas and oxygen alternately over the small and large electrodes. This led to the testing of a cell having a silver-oxygen-carbon dioxide cathode and a platinum-hydrogen anode. I n Figure 2, curves 1 and 2 for hydrogen-oxygen plus carbon dioxide galvanic couples show that for current densities up to 150 ma. per sq. cm. at 500" C. no polarization occurs other than the purely ohmic drop. T h e correction for the ohmic cell resistance, as determined with alternating current (1000 c.P.s.), changes these curves to the nearly horizontal lines, 1' and 2'. Similar experiments with air and carbon dioxide (oxygen pressure, 0.17 atm.) have shown that under these circumstances, polarization of the silver electrodes at 500' C. is negligible also. Thus, silver powder is an ideal oxygenFuel
Ha
EO at 800° K.,Volts EO at 1000° IC., Volts
1.05 1.00
+
C (to '202) 1.03 1.03
carbon dioxide electrode in any carbonate fuel cell operating a t 500' C. or higher. Curves 3 and 3' show the necessity of a cathodic carbon dioxide supply, and curves 4 and 4' reveal polarization of the platinum electrode with carbon monoxide a t 600' C. When it was found that the silver-oxygen electrode does not polarize, various anodic metal powders were studied to determine their usefulness as hydrogen, carbon monoxide, and methane electrodes. For hydrogen, platinum and nickel seem
304
suitable. For carbon monoxide, galvanic activity at 700' C. of the following metals was Pt > platinized Fe or Ni > Fe > Ni > Co > Cu > Cr > Mn For this series, however, differences in particle size may be important, and the general reproducibility of results with similarly constructed cells has not been too good. Moreover, above 750' C., differences in performance become negligible. For methane, no satisfactory electrode metal has been found for use below 750" C., but if steam is added, nickel seems suitable. This is probably because nickel catalyzes formation of hydrogen and carbon monoxide : CH, H20 % CO 3H2 (1) Theoretically, equilibrium reactions such as Reaction 1 do not cause important losses in potential electrical energy because the associated free enthalpy change is small compared to that of the ideal galvanic combustion of methane with oxygen. This can be seen by comparing the standard electromotive forces (in reference to oxygen) of different fuel gases : GO 1.10 1.01
+
CHI 1.04 1.04
+ HzO % COz + Hz
(2)
Hydrogen is by far the most galvanically reactive gas and, moreover, its diffusion constant is 5 to 8 times greater than that of carbon monoxide or methane. Therefore, in mixtures of methane, car-
INDUSTRIAL AND ENGINEERING CHEMISTRY
Y
I
c l
C9He 1.07 1.08
Since at 1000' K., EO for both hydrogen and carbon monoxide is only slightly smaller than that for methane, losses in ideal electric energy, associated with Reaction i , are small. Similar arguments hold for CO
bon monoxide, and hydrogen, the principle electrochemical reaction is expected to be oxidation of hydrogen to steam. Consequently, equilibrium 1 or 2 will shift to the right to form fresh hydrogen. The essential kinetic problem, therefore, is to ensure that the rate of this shift can keep up with the galvanic conversion rate for the hydrogen. Because Reactions 1 and 2 may proceed over the full electrode surface, there is no fundamental reason why methane cannot be burned with sufficient velocity. The electrode in this case should be both a good methane-steam reforming catalyst and a good hydrogen electrode. Nickel powder is fairly satisfactory for this double function. For measuring cell performance, only stationary values have real significance. Since the partial pressures of reaction products increase with increasing conversion of the fuel gases, the terminal voltage is a function of both the gas feed rate and current density when the gas supply is continuous. In experiments described in the literature, gas feed rates nearly always seem substantially larger than their consumption rates which are proportional to the current drawn. When leakage occurs across the electrolyte, this excess is necessary to flush the electrodes, but it means that reaction products cannot accumulate. If fuel gases, practically free of reaction products are fed into the cell, data thus obtained give no more than a snapshot of optimal operating conditions. Voltage efficiencies derived in this way overestimate real performance because conversion of fuel gas must be at least 80% for practical use of a cell battery. This is illustrated in Figure 3 where, because of increasing polarization at increasing conversion A and current density i, actual voltage curves fall more rapidly than the theoretical cell voltage
1
1
0
A,
I
1
1
AI
u 1
Figure 3. Terminal voltage as a function of current density and conversion. Because of polarization not all gases can be converted. True voltage efficiency i s the ratio of the areas under the V t and E curve, bordered b y the X axis, p and q. The overestimates cell ratio, V,(Xl)/E(Xl), performance
---
-
average terminal voltage, Vt, at
i = ia
FUEL CELLS
20 o / o co 60 O/o C02 20°/0 H2O 3'/0 02 ~~
10
0
20
30
40
60
50
-
70
mA/cm2
Figure 4. These curves are measured at optimal conditions. To avoid leakage, oxygen content of cathodic gas is adapted to the carbon monoxide content of the anodic gas
Vt; therefore, complete conversion a t nonzero current density cannot be reached. The voltage efficiency thus described corresponds to the ratio V J E at X = XI; however a more realistic measure is the ratio
bdX-'.g.,
Over
O n nickel powder anodes a t 600" to 650' C. and with air and carbon dioxide as cathodic gas passed over silver powder electrodes, hydrogen containing up to 50 mole % of water can deliver current densities of 50 ma. per sq. cm. at 0.80 to 0.70 volt, and 100 ma. at 0.60 to 0.40 volt. Such cells can operate for several months at 650' C. undkr steady conditions with only a slow electrolyte loss which causes a corresponding decrease in terminal voltage and an increase in internal resistance. Polarization, other than purely ohmic drops, seems negligible. Methane is more difficult to handle, but nickel electrodes at 750" C. or higher are promising if steam is added. At 770' C. methane containing 30 mole yo of water gave the following results:
the
J
J
range of X I to hz. T o obtain results of this kind, a number of characteristics can be measured, using gases of different A's a t relatively rapid feed rates. Such results correspond to the behavior of a series-connected battery where only a small fraction of fuel gas is oxidized in each cell. Therefore, gas feed rates considerably greater than the electrochemical rate of turnover were used for the experiments reported in this article.
Current dens., ma./sq. cm. Terminal voltage, volts
0 0.98
40 0.63
20 0.80
60 0.50
0
Table I. Voltage and Thermal Efficiencies of Galvanic Combustion of Carbon Monoxide a t 720' C. (From Figure 4; fuel gas, CO C o t ; xz = 0.75 mole CO; 1-5; = 0.25 mole CO?; xf = moles CO that cannot be converted) 20 i MA./sq. cm. 10 30 40
+
Pt, volts co
llvolt
lltherm
co
xr Oia
qtherm
carbon"
Integrated system.
-
0.79 0.78 0.54
0.72 0.71 0.49
p = O
p = l
p = o
p = l
0.05 0.36 0.75
0.08
0.10 0.34 0.66
0.16 0.59 0.60
0.61 0.71
0.65 0.64 0.44 p
0 p = l
0.15 0.32 0.65
0.23 0.56 0.55
0.60 0.59 0.41 p = o
0.20 0.30 0.52
p = l
0.24 0.52 0.44
For mixtures of carbon monoxide and dioxide at 650" to 700' C., iron or ironnickel electrodes are sensitive to small leakages of oxygen from the cathodic side, especially when the carbon dioxide content is over 5.0 mole %. The open circuit voltage decreases considerably below the reversible value. Conversely, the silveroxygen-carbon dioxide electrode is less sensitive to carbon monoxide leakage, probably because it has excellent cathodic activity. Performance for mixtures of carbon monoxide and dioxide improve considerably if oxygen in the cathodic gas is reduced to a few volume per cent. By means of this artifice, the characteristics of Figure 4 can be obtained. At 720' C. and 10 ma. per sq. cm. these mixtures can be burned to a final carbon monoxide content of 5 mole yo; at 21 ma. they can be burned to 10 mole yo. Assuming that a mixture of x , moles of carbon monoxide and 1 - x , moles of carbon dioxide enters a series-connected battery, and that x j moles of carbon monoxide escape in the outlet gases, the quantity of carbon dioxide in the latter gases is (1 - x , ) (1 PI( x , - XI) when P is moles of cathodic carbon dioxide needed per mole of anodic carbon monoxide to prevent excessive polarization. For strictly reversible cells, cp equals 0; for the most unfavorable case it equals 1. In practice 0 < cp < 1, depending on diffusivity of carbon dioxide across the electrolyte, for which no exact data are known. Assuming that rp = 1, the carbon monoxide fraction in the outlet gases is
+ +
Yco =
XfA1
+ xz - x r )
If p equals 0, then ycoequals x I . The average voltage, associated with galvanic combustion of carbon monoxide
is
When the curves of Figure 4 are translated to curves of V , us. ( x , - x i ) which use current density as a parameter, as in Figure 3, this voltage can be found by graphical integration. Voltage efficiency for carbon monoxide combustion may now be defined as vvoltage= Vt/Eo where Eo is the standard electromotive force for the carbon monoxideoxygen couple at the operating temperature-namely, 1.01 volts at 720' C. Here, carbon monoxide which escapes with the outlet gas is ignored. However, in an integrated fuel cellgasification system, recycling part of the the outlet gas over coal or coke is the most important advantage of energy-delivering cells. VOL. 52, NO. 4
APRIL 1960
305
In Table I, data in the two bottom rows refer to a process where a fraction (1 - cy) of the anodic outlet gas is recycled over carbon to recover the original mixture of carbon monoxide and dioxide as anodic fuel gas. Here, thermal efficiency is based on heat value of the carbon; heat value of the carbon monoxide in the waste fraction cy is considered a total loss. Clearly, a promising characteristic such as 0.8 volt a t 50 ma. per sq. cm. for 1 0 0 ~ o carbon monoxide in Figure 4 probably indicates little about the actual behavior of carbon monoxide. Considerable improvement over the data reported can be expected for cells that are gas-tight and operate on greater oxygen concentrations.
Literature Cited
( I ) Davtyan, 0. , K., Bull. mad. sci. U.R.S.S. Classe scz. tech. 1946, pp. 107, 205. RECEIVED for review September 15, 1959 ACCEPTED December 31, 1959 Division of Gas and Fuel Chemistry, 136th Meeting, Atlantic City, N. J., September 13-18, 1959. Based on work done at the Central Technical Institute T.N.O. in partial fulfillment of requirements for the degree of doctor of philosophy.
G. H. J. EROERS and J. A. A. KETELAAR Central Technical Institute T.N.O., The Hague, Netherlands
Nature of the Electrode Process
HI,,
TEMPERATURE CELLS operate a t temperatures ranging from 500O to 900' C., and are unique in their ability to use carbon-containing fuel gases The most successful results have been obtained with electrolytes of fused alkali carbonates ( 7-3)-the electrolyte is not contaminated with carbon dioxide and other carbonaceous products. Fuels are probably used indirectly through reforming and thermal cracking reactions which can be carried out internally or externally to the cell operation. Hydrogen is produced thereby which readily reacts at the fuel electrode. Carbon monoxide may also be consumed directly at the fuel electrode but does not seem to be as active. I t may be consumed indirectly by conversion to
t
A
hydrogen within the cell by reaction Tvith steam which aids considerably in its utilization ( 7 ) . Thus. behavior of the hydrogen and oxygen electrodes is controlling for good performance, and the study reported here is concerned with the reaction mechanism of these electrodes. The cell used was similar to that of Broers ( I ) , and the electrolyte was mixed alkali carbonates disposed on a specially prepared pure porous magnesia matrix. Operating data were obtained over the range of 700' to 800' C., using both hydrogen-steam mixtures and carbon monoxide-dioxide mixtures a t the anode and air a t the cathode. Porous sintered nickel and iron (average particle size, 65 microns) were used as fuel electrodes
T--7-q - -
-_I -1 LL Wd = 0 0111 SQUARE MESH ELECTRODE ~~
PARALLEL WIRE ELECTRODE
40
Figure 1. Effective resistance ratio R , f f / R depends on the ratio between electrolyte thickness and contact spacing ( 1 /d). Fractional area covered i s (A/d)2for a square mesh electrode and (A/d) for a parallel wire electrode
2
3
4
5
678910
20
30
l/d
306
INDUSTRIAL AND ENGINEERING CHEMISTRY
40 50 6 7 8 9100
while both silver gauze (90 mesh) and lithiated sintered nickel oxide (4) were used a t the air electrode. T h e voltage current characteristics were analyzed statistically. After taking into account change in gas composition caused by cell reactions, drop in voltage was linear with current drain. T h e linear relationship held within about 0.06 volt as determined by the 95% confidence limits. T h e measured specific internal resistance was rather high, howeverabout 4 to 7.0 ohm-cm. The calculated internal resistance from voltage drop during current drain was in some cases equal to the measured resistance while in others it was u p to 40% higher. The high internal resistance limits the cell output and is in the same range as observed by other workers. Experimental work has shown that this is not caused by some peculiar property of the electrolyte matrix. T h e internal resistance of the matrix was measured when fully loaded with melt and pressed between two flat silver gaskets as electrodes. An average value of 0.7 ohm for resistance per sq. cm. was found in the temperature range of 700' to 800' C. This is smaller by a factor of 7 to 10 than that observed during cell operation. This phenomenon can be explained: For the gas to have proper access to the electrode surface, the melt inventory must be adjusted until a small area of contact is maintained between electrode and electrolyte. Theoretical calculations were made to estimate the actual contact area for several geometrical arrangements by solving the Laplace equation,
with appropriate boundary conditions. T h e calculated resistance for a given contact area depends strongly on geometrical arrangement and spacing of the contact points (Figure 1). T h e calculations show, however, that the actual fractional contact area must be very low to explain the observed high resistance and probably no higher than 3x This figure corresponds to an estimated value of l / d for the silver gauze electrode of 6.3. Some mechanism must be operating to broaden the three-phase limit where electrode, electrolyte and gas meet. Otherwise two deleterious factors become effective-i.e., activation polarization as a result of concentrating the electrode reaction on a very small area, and the concomitant high effective resistance. The two most likely mechanisms are permeation of the gas through the bulk electrode metal, and diffusion across the electrode surface. A mathematical treatment has been worked out