/III. .I Il~~‘llYJ&wl Flwrg~~, Vol 2 I. No 2. pp.

x7 93, 199h
Pergamon Copyright @ International Associnrlon for Hydrogen Energy
Elscwzr Scicncc Ltd
Printed m Great Rntam All right\ rcscrved
0360-3199(95)00005-4 mm-3 199.9h 5 15.00 + 0.00

DES I GN AND PERFORMANCE OF A SOLID POLYMER ELECTROLYTE

WATER ELECTROLYZER

P. MILLET.* F. ANDOLFATTOt and R. DURAND;

* Laboratoire des Composts non Stocchiometriqucs, Universitt- de Paris Sud, Bdt. 415. 91405 Orsay. France
t Laboratoire d’Applicalions Sp&alc~ de la Physique, Centre d’Eludes NuclCaires de Grenoble, 38041 Grenoble, France
z Centre dc Rechcrche en Electrochimie MinCrale et en CZnie des Pro&d&, Domaine Universltaire.
38402 Saint Martin d’Hires. France.

Abstract -The development of medium size solid polymer electrolyte (SPE) water electrolqzers is of great practical
interest for on-board production of oxygen in submarines and for energy storage purposes in space applications.
While the SPE technology has largely demonstrated its potential on the laboratory scale, very little information is
available in the literature on the technological characteristics and electrochemical performances of larger water
electrolyzers. In this paper, we present part of the experience that we gained in the development of medium si7e
electrolysis units. A 0.5 kW SPE water electrolyzcr was designed and tested. Its performance is discussedwith respect
to technological design (fluids and current distribution) and operating conditions.

INTRODUCTION Ohmic drop across the membrane is minimized when

using protonic conductor materials. NaGon is the typical
Solid polymer electrolyte (SPE)g water electrolysis tech-
membrane material used for such applications. In Nafion
nology is frequently presented in the literature as a very
membranes, the acidity is similar to that of a 20 wt%
interesting alternative to the more conventional alkaline
sulfuric acid solution. For this reason, acid resistant noble
water electrolysis. metal catalysts (such as platinum and iridium) or their
To date, however, little work has been published
oxides must be used for electrocatalysts. The main
regarding the problems which may be encountered in the drawback which can be argued against this technology
development of SPE electrolyzers. Most of the work done is its expense. What is gained with the improved power
in the field has to be credited to the General Electric
efficiency is lost with the material cost. The cost of the
Company (U.S.A.) who first proposed the concept of SPE
electrodes added to the cost of the solid electrolyte
tehnology for the development of hydrogen/oxygen fuel
seriously limits the development of SPE water electro-
cells [I- 61. More recently, the Swiss firm ABB developed
lyzers in the industry. However, when cost consideration
its MEMBREL process both for water electrolysis and is of less importance-such cases do exist-SPE technol-
ozone generation [7-91. Japan is also active in this field
ogy has to be considered with great care. Consider, for
with the work performed under the Sunshine project
example, oxygen generation in submarine crafts where
[lo- 121.
security and reliability are particularly crucial. Raw
The inherent advantages of SPE technology over the
material (water) is taken from the surroundings of the
alkaline one are clearly established: (i) greater safety and
vessel and treated to eliminate organic, biological and
reliability are expected since no caustic clcctrolytc is
chemical undesirable products. While oxygen is used for
circulated in the cell stack; (ii) tests made on bare
human breathing, the hydrogen by-product resulting
membranes have revealed that some materials could
from water splitting has to be eliminated. This can be
sustain high dilTcrentia1 pressure without damage [13] done in two diIIerent ways: (i) storage under pressure or
and were efficient in preventing gas mixing; and (iii) the in a suitable absorption medium (hydride, etc.); and (ii)
possibility of operating cells up to several amps per square
throwing out at sea. The first possiblity is usually rejected
centimeter with typical thicknesses of a few millimeters because of safety considerations. The second requires the
is afforded [14]. use of noisy and polluting gas compressors to overcome
the external pressure. For such reasons, technology
presenting high security and reliability and allowing high
2jSPtz IS a trademark of HamIlton Standard, W3.A. rating pressure (100 bar and above) is valuable.

RI
88 P. MILLET et ul

(4 (b)

Membrane thickness

Fig. 1. Metallic platinum concentration profile across the membrane thickness (e) obtained by electron microprobe analysis. (a)
Nation-platinum exchanged sample. (b) Same sample treated with a 16 g I-’ NaBH, solution for 2 h. (---) Pt” and (-) Na+
profile.

During the last 2 decades, SPE technology has been 25°C

2.5 -
the subject of intensive research in Grenoble, for water
2.4 -
45°C
electrolysis, hydrogen/oxygen and alcohol fuel cells 55°C
applications. Only water electrolysis will be considered 70°C
in this paper. Platinum and iridium based efficient 80°C
electrode membrane electrode (EME) composites have
been developed and tested [14, 151. Laboratory scale
experiments have been performed over more than 25,000
h of continuous electrolysis for some samples. They all
showed stable cell voltages during the tests and u
posteriori analysis of the electrode surfaces has shown
that no deterioration had occurred [16]. 1.6l I I I I I I
The next stage was to become more involved in the 0 200 400 600 800 loo0 1200
technological development of SPE water electrolyzers. Current density (mA cmM2)
This has been done over the past year and part of the Fig. 2. Cell voltage vs current density obtained for a 5 cm*
results obtained are the subject of this paper. Pt/Nafion/Pt EME composite. Effect of the operating temper-
ature.
PLATING PROCEDURE
Platinum based composites
1). Several loading-precipitation cycles can be performed
Details of the plating procedure used for the prepara- to increase the amount of metallic platinum deposited.
tion of platinum based EME composite have been Improved cell performances can be obtained by etching
described elsewhere [15]. Briefly, metallic platinum is the membrane surface and electroplating iridium on the
plated to the membrane surface according to a two-step anodic side [16]. Membrane etching using ionic abrasion
process: (i) an ion-exchange process during which the techniques should, however, be used with great care, as
membrane, initially in H+ form, incorporates a given the energetic gain obtained by the diminution of the
amount of a cationic precursor salt [Pt(NH,),]” [17]; membrane thickness is balanced by lower mechanical
(ii) the in situ precipitation of the salt to the metallic state stability and higher gas permeability. Figure 2 shows the
within the membrane by using an aqueous solution of typical current density vs potential curves obtained with
sodium borohydride. Under adequate preparative con- such EME composites at different operating tempera-
ditions, precipitation of metallic platinum occurs into the tures. Long time tests performed on such EME compos-
membrane, preferentially at the membrane surfaces (Fig. ites have shown an excellent stability of the cell voltage
 AN SPE WATER ELECTROLYZER

Time (hours)
Fig. 3. Time varation of the cell voltage of a 2.3cm’ Pt/Nafion/Pt
EME composite operated at 0.5 A cm’, atmospheric pressure Fig. 4. Plexiglass cell used for the preparation of IO0 cm2 Pt
and room temperature. Values of the cell voltage were not based EME composites.
corrected for the ohmic drop across the current collectors.

(Fig. 3). Such performances can be obtained if and only composites have been determined and models have been
if the platinum deposits meet the following conditions: developed to optimize the localization of the platinum
(1) The metallic electrode must be located not too far particles and to improve the electrochemical properties
inside the membrane, otherwise faradaic yield will be of the EME composites [lS. 191.
decreased because of gas diffusion across the membrane. The same technique was used for the preparation of
(2) The metallic electrode must be located simulta- 50 and 100 cm2 EME composites using an adequate
neously inside the membrane (where it acts as an elec- plating cell (Fig. 4). With typical platinum loadings of 1
trocatalyst in contact with the solid electrolyte) and mg cm2 per face it was possible to reproduce the
outside, on the surface of the membrane (where it acts electrochemical performances of laboratory-scale
as current collector). samples (l--5 cm’). The plating process was found to be
This is achieved by monitoring the plating conditions, appropriate for the preparation of large EME composites.
namely the amount of platinum tetramine which incor-
porates the membrane during the ion-exchange, and the Iridium oxide based cells
concentration of the reducing sodium borohydride. The
Iridium oxide is also an interesting material both for
main parameters controlling the preparation of the EME
oxygen and hydrogen evolution. lridium oxide was

*.J .... ......I......................................~~.~.~~~....~_....-....-..........-....--....-................-..................--...... 1

2.2 -------- ........ ... ... .. ... .... . ......... ... ......... ........ ...-....-...__-........^... .. . .
3 -
1,* i_T---”
- .......... .. .. .... .... ..... .... .. ... .... .. ..... ..... ....... ..... .. .. .. ..... ...... ... .....

.....__......._..................................~.....
_......- .......~...1...._......___..................................,

l,o- _.............. .. .. .. . ...... ....... ... .

f :;-:sQ ;;;;;I: ~111IIIIIIIII1:11::III:~

3
u 0.2__.._..l......................................... . i = 1oo mA,cmZ . .. ... ... ............ .... .... _.....-. ..... ... .. .... ..
_ ______..._..
__..___......._.........................~ ...__._ ... .. ... ..... ....... .. ... ... .........- ......1.
4,2- ......._.........._..................................-............................................................................._................
--%I PI
-0.6 .. ....._.._._............................................................................................
-+ J-Jw.4
.. . . .....

-1.0 I
0 loo0 2ooo 3ofxJ 4ooo 5GQa 6ooo 7ooo
Time (hours)

Fig. 5. Time variation of the cell voltage, anodic and cathodic overvoltages, for a 5 cm’ Ti/IrO,/Nafion/Ti EME composite
operated at 0.5 A cm’, room temperature and atmospheric pressure.
 P. MILLET rt al.

Fig. 7. Photograph showing the details of the cell stack

The main problem encountered in developing this

technique was the formation of insulating TiO, during
the thermal formation of the iridium oxide. Ionic implen-
tation of noble metals may be an interesting solution to
overcome this problem.

ELECTROLYZER DESIGN
Generd charucteristics
The electrolyzer is of the filter press type and has the
following characteristics (Figs 6 and 7).
(i) Cell characteristics:

Fig. 6. Photograph of a two-cell electrolysis stack in its pressure l membrane, Nafion 117 (1100 equivalent weight; dry
vessel: (a) anodic side; (b) cathodic side. thickness, 0.178 mm);

thermally deposited directly on the porous titanium foil

as follows [20]. IrCI, was dissolved, either in water or
in ethanol, and the solution was dried until a paste was
obtained. The paste was then applied onto the substrate.
A pre-heating treatment was performed at medium tem-
perature (100°C for 1 h) to eliminate the remaining
solvant and was followed by a thermal annealing treat-
ment at 600°C for 5 h.
Iridium oxide presents a number of advantages by
comparison with metallic platinum: (i) the anodic over-
voltage is significantly reduced (100 mV at 1 A cm-‘);
(ii) the cathodic overvoltage is increased (50 mV at 1 A
cm-‘), but more stable results are obtained with time.
This is because platinum is highly sensitive to the presence
of traces of ions such as Cu’+, Ni2+, etc. which may be
produced by the slow but steady corrosion of the steel
pipe of the test bench. This may result in a very significant
increase of the cathodic overvoltage by under-potential
deposition (UPD) of these cations at the platinum surface.
Figure 5 shows the time variation of the cell voltage,
anodic and cathodic overvoltages at 500 mA cm-2 Fig. 8. Plexiglass cell used to measure the load losses across the
obtained with a Ti/IrO,/Nafion/Ti EME composite. cell stack.
 AN SPE WATER ELECTROLYZER 91

Wuter distribution und gas collection

Water distribution on both anodic and cathodic sides
was achieved by forced circulation of deionized (18 Ma
cm) water in a machine-madegrooved titanium separator.
Water was circulated through the grooves for reaction
feeding, heat removal and gas transport to the gas-water
separation unit (Fig. 8). Friction losses were controlled
(Fig. 9) by adjusting the diameter of the inlet channel to
ensure homogeneous water distribution in each compart-
ment of the cell.
During electrolysis, an extra amount of water is
transferred to the cathodic side by electro-osmosis (ap-
proximately 3.5 water molecules per proton). This transfer
was balanced by applying a small differential pressure
between the two compartments.
The EME composite was mounted on a polysulfone
frame (Fig. 10). It is vital that the membrane remains
hydrated, both during electrolysis and at rest, otherwise
it shrinks and tears out of the frame.

Current distribution
9. Friction losses within a unit cell as a function of the
diameter of the input channel. Current distribution perpendicular to the EME com-
posites was made as follows. The cells were separated by
the grooved titanium plates. Two thin (1 mm thick)
porous titanium sheets were mounted on each side of the
l cathodic catalyst, platinum; EME composites to ensure a homogeneous current
l anodic catalyst, platinum; distribution at the electrode surface (Fig. 11). These sheets
l cell area, 0.01 m’; were obtained by sintering and porosities as high as 50%
l current density, 10 kA rn- ‘; were obtained. Such porosities are necessary, especially
l cell voltage, 2.2 V at 1 A cm -’ and 80’ C; on the anodic side of the cells where two opposite mass
l oxygen generation rate, 0.5 m3 day- ‘. flows take place pcrpcndicular to the membrane sufacc:
water from the distributor to the membrane and oxygen
(ii) Stack characteristics: from the membrane to the collector. The diameter of the
oxygen bubbles being inverserly proportionnal to the
l number of cells by stack, 2; rating prcssurc, improved clcctrochcmical proformances
l electrical input power, 0.5 kW; are generally obtained at elevated operating pressure.
l operating pressure range, 1 100 bars; Other materials were tested as cathodic current dis-
0 operating temperature range, 7OGlOO”C tributors such as carbon foil or nickel foam but the best
l stack volume, 5.3 drn’. results were obtained with titanium based distributors.

Fig. 10. Schematic representation of the fixation of the EME composite on the polysulfone frame. (1) polysulfonc frame; (2)
circulation channel: (3) inlet channel; (4) grooved titanium separation; (5) water distribution groove; (6) polysulfone fixator; (4)
titanium current disrributors; (8) EM!? composite.
92 P. MILLET et al.

,I 0.25 0.50 0.75 1.00 125

Current density (A cme2)
Fig. 13. Cell voltage vs current density obtained at different
temperatures with a 50 cm2 Pt/Nafion/Pt EME composite.
Dashed line: laboratory scale results obtained at 8O’C.
Fig. 11. Detail of a porous titanium current distributor.

Test bench lyzers of up to 100 kW input power. Deionized water at

The electrolyzer was tested using the bench schemati- room temperature was introduced to the pressure vessel
zed in Fig. 12. This bench was designed to test electro- where it was pre-heated (cooling of the stack). Water then

02H*
entered the anodic loop to feed the reaction. For larger
electrolyzers, a heat exchanger could be activated to
maintain the operating temperature at the desired value.

9
ELECTROLYZER PERFORMANCES

One-cell conjguration

ki
The current densities vs potential relations obtained
LL
at various operating temperatures with a 50 cm2 Pt/
Nafion/Pt EME composite clamped between two porous
P P titanium sheets are shown in Fig. 13. The dashed curve
concerns the results obtained at 75°C at the laboratory
scale and is given for comparison.

Two-cell conjiguration
Results obtained with a two-cell configuration are

I
1 given in Fig. 14.Again, laboratory scaleperformances are
given for comparison.

8
E (b)* * * *
3aI6 * *
B * * 0 0 0 O
g4 * E 0 O O (Z)
I I I I I I
21
0 0.2 0.4 0.6 0.8 1.0
Currentdensity (A cm-*)
Fig. 14. (b) Cell voltage vs current density obtained at room
Fig. 12. Schematic diagram of the bench used to test the temperature with two 100 cm’ Pt/Nafion/Pt EME composites.
electrolyzer. (a) Laboratory scale results obtained at 80°C.
 AN SPE WATER ELECTROLYZER 93

The slope of curve (b) is 30 mV decade-’ (10 mA flows were observed from one compartment to the other
cm-‘). Correction is made for the ohmic drop across the with differential pressure as low as 20 mbar. At high
membranes; an ohmic term of 26 mV decade- r remains, operating pressure, a severe control of the differential
which can be attributed to the bad electrical contacts in pressure across the cells is required.
the cell stack. Ideally. for such configurations, an ohmic Corrosion of the steel pipes was found to be another
term of 1 mV should be obtained. This would result in crucial problem which has to be considered with great
a 100 mV ohmic loss at 1 A cm-‘. These figures highlight attention to avoid a drastic increase of the cell voltage
the problems encountered in the development of filter during operation.
press electrolyzers; however, it should be kept in mind
that ohmic losses are not necessarily undesired. Heat
REFERENCES
generated by the ohmic term is transferred to the feed
water circulating in the electrolyzer and allows operation I. W. T. Grubb Jr, J. Electrochem. Sot. 106, 275 (1959).
of the stack at the desired operating temperature. 2. W. T. Grubb Jr. US Patent No. 2.913.511. 17 November
Another important drawback observed during the run (1959).
3. R. H. Blackmer. US Patent No. 3,061,658(30October 1962).
tests was the slow but steady increase of the cell voltage.
4. W. A. Titterington and J. F. Austin, Abstract 233, p. 576,
The reason was found to be due to the corrosion of the The Electrochemicul Society E.uended Ahsiroct, Vol. 74--2,
stainless steel pipes of the manifolds contacting the bigly New York, 13-17 October (1974).
deionized (18 MR cm) water. While Fexf, Nix+ and Crx+ 5. P. W. Lu and S. Srinivasan, J. Appl. Ekectrochem. 9, 269
concentrations in the water feeding remained low, those (1979).
cationic impurities were trapped and concentrated in the 6. Solid polymer electrolyte water electrolysis technology de-
membranes. This resulted in an increase of the ohmic velopment for large scale hydrogen production, General
drop in the membrane as well as an increase in the Electric Company, DOE Report DOE/ET/26 202-I (1981).
electrode overvoltages. On-line deionizers give more 7. S.Stucki and R. Muller, Proc. Third World Hydrogen Energy
Cor$ ( WHEC’), Tokyo, Vol. 4, p. 1799 (1980).
stable cell voltage but are a limitation for operating at
8. H. Nagel and S. Stucki. US Patent No. 4. 326,930 (27 April
high pressure and elevated temperature. 1982).
9. R. Oberlin and M. Fischer, Proc. Sixth WHEC, Vienna, Vol.
CONCLUSIONS 1. p. 333 (1986).
10. H. Takenaka, E. Torikai, Y. Kawami and N. Wakabayashi,
An SPE water electrolyzer of 0.5 kW input power was Inr. J. Hydrogen Energ), 5, 397 (1982).
designed and tested. The plating procedure used at the 1 I. A. Katayama-Akamata, H. Nakajima, K. Fujikawa and H.
laboratory-scale for the preparation of l-5 cm2 EME Kita, Ekctrochim. Acta 28, 777 (1983).
composites was used for the preparation of larger com- 12. T. Ohta and 1. Abe. Proc. F$h WHEC, Toronto, Vol. I. p.
47 (1984).
posites (100 cm’). Homogeneous platinum deposits and
13. A. B. Laconti, E. N. Balko. T. G. Coker and A. G. Fragala,
similar electrochemical performances were obtained. Proc. Electrochemicul Socier c’ Srm. on Ion- Exchange. Trans-
The main problem connected to the operation of the port und Inter$xiul Properties, Vol. 2, p. 318 (1981).
electrolyzer was the presence of substantial ohmic losses. 14. P. Millet, M. Pin&i and R. Durand, J. Appl. Electrochem.
At high rating pressures, gaseous production accumulate 19, 162 (1989).
between the electrodes and their current collectors and, 15. P. Millet, R. Durand and M. Pin&i, Inr. J. Hydrogen Energ)
as a consequence. important ohmic losses appear. Non- 15, 245 (1990).
homogeneous current distribution at the electrode---solid 16. P. Millet, T. Alleau and R. Durand. J. Appl. Electrochem.
electrolyte interface results in the formation of hot spots 23, 322 (1993).
17. P. Millet, R. Durand, E. Dartyge, G. Tourillon and A.
which can eventually lead to the melting of the membrane.
Fontaine, J. Elecrrochem. Sot. 140, 1373 (1993).
Special attention must be given to the design of the 18. P. Millet, F. Andolfattoand R. Durand,J. Appl. Electrochem.
current distributors to avoid such drawbacks. 25, 227 ( 1995).
Rather surprisingly, the EME composites were found 19. P. Millet, F. Andolfattoand R. Durand, J. Appi. Elecrrot6em.
to be very sensitive to the differential pressure between 25, 233 (1995).
the anodic and cathodic compartment. Significant water 20. F. Andolfatto and R. Durand, Electrochim. Actcr(submitted).