Journal of Power Sources 136 (2004) 232–239

CLIPPER: a long-range, autonomous underwater vehicle using

magnesium fuel and oxygen from the sea
Øistein Hasvold a,∗ , Torleif Lian a , Erik Haakaas a , Nils Størkersen a ,
Olivier Perelman b , Stephane Cordier b
a FFI (Norwegian Defence Research Establishment), P.O. Box 25, N-2027 Kjeller, Norway
b Direction General d’Armament/Bassin d’èssais des carènes (DGA/BEC), Chaussée du Vexin, Val de Reuil 27100, France

Available online 8 June 2004

Abstract

The development of the CLIPPER autonomous underwater vehicle (AUV) is a joint research project (JRP) between DGA/BEC in
France and FFI in Norway. The power source is a semi-fuel cell using oxygen dissolved in seawater as oxidant, seawater as electrolyte
and magnesium as fuel. Predicted range for the vehicle is in excess of 1600 nautical miles at a speed of 2 m/s. Design depth is 600 m. This
long endurance, as well as the large internal volume of 740 l of the pressure hull, makes the CLIPPER vehicle an excellent tool for long
range subsea operations.
© 2004 Elsevier B.V. All rights reserved.

Keywords: Magnesium anode; Oxygen cathode; Seawater electrolyte semi-fuel cells; Magnesium seawater batteries

1. Introduction namic optimisation and testing and propulsion technology,

while FFIs contribution is within power sources and AUV
The first autonomous underwater vehicle (AUV) devel- technology. Corps Laminaire Integrant une Propulsion a
oped by FFI was a technology demonstrator made to show Pile Eau du meR (CLIPPER). The target for the first phase
that that the magnesium/dissolved oxygen semi fuel cell of the CLIPPER project is to build a full-scale model vehi-
[1] could be used as a power source for AUVs. This was cle and power source and verify the battery and propulsion
successfully demonstrated in May 1993, but these experi- performance in full-scale tests. One of the objectives of the
ments also highlighted the low specific power of this power JRP is that the model vehicle should be capable of being
source, making it inferior to other power sources unless further developed to an ocean going prototype to demon-
long-term missions (>200 h) were a requirement. Thus later strate potential operational capabilities in a follow-on phase
AUV developments from FFI (e.g. the HUGIN series) have 2 of the CLIPPER collaboration. Thus the filament wound
been using other power sources such as the alkaline alu- pressure hull in the model vehicle is designed for an oper-
minium/hydrogen peroxide semi-fuel cell [2,3] and lithium ational depth of 600 m.
ion batteries [4]. However, compared to the seawater battery, Prospective military applications of CLIPPER that exploit
these chemistries have low specific energy. the extreme range capability or endurance are intelligence,
The aim of the CLIPPER project was to establish a tech- surveillance, target acquisition, reconnaissance (ISTAR) op-
nology base for providing the capability of very long-range erations, anti submarine warfare (ASW), offboard tactical
operation of AUVs by combining optimal hydrodynamic element for submarine operations, and rapid environmen-
design and efficient propulsion with high energy density bat- tal assessment (REA). Marine research, polar research and
tery technology. The project is organised as a joint research search and survey operation are among some civilian appli-
project (JRP) between France and Norway. The partners cations that may use a range capability in excess of 1600
are Direction Generale Armament (DGA), Bassin d’Essais nautical miles.
des Carenés (BEC) and the Norwegian Defence Research Because of the intimate connection between the vehicle
Establishment (FFI). BEC provides expertise on hydrody- drag and propulsion, the internal flow, the battery perfor-
mance and the flow through the battery, the development has
∗ Corresponding author. Tel.: +47-63-80-75-34; fax: +47-63-80-75-09. consisted of a series of loops. In each loop a vehicle was
E-mail address: oistein.hasvold@ffi.no (Ø. Hasvold). designed and power for propulsion and flow in the battery

0378-7753/$ – see front matter © 2004 Elsevier B.V. All rights reserved.
doi:10.1016/j.jpowsour.2004.03.023
 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239 233

Fig. 1. Schematic drawing of the CLIPPER AUV. Displacement  = 3.9 m3 .

Fig. 2. Seawater flow through the CLIPPER AUV.

computed. A model battery was then produced and tested vantage of this concept is the increased ratio of energy to
at the calculated flow velocity and specific load. Based on weight and volume compared to common battery technol-
the test results, a new vehicle and battery design was devel- ogy. A further weight saving is caused by the operation at
oped. The final design is shown in Fig. 1. Fig. 2 shows the ambient pressure. Most other battery systems require a pres-
flow of water through the vehicle. Thrust for propulsion is sure resistant container for the battery.
generated partly from the water from the battery (water jet), During operation, a flow of seawater through the battery
partly from the propeller. supplies the battery with oxygen and removes the products
of the reactions taking place in the battery. Comprehensive
descriptions of the cell chemistry are given in [1,5,6,7].
2. Principles of operation One challenge in the design of the battery was to be able
to supply the battery with a sufficiently large flow of seawa-
The battery concept is based on the use of seawater as ter and at the same time use as little energy for this process
electrolyte and oxygen dissolved in the ocean as oxidant in as possible. A second challenge is to handle large currents
a semi-fuel cell with magnesium anodes as fuel. The ad- with small voltage loss. Because the cells are in a common
234 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239

electrolyte (the ocean), serial connection of seawater cells to lost measured. In this paper, only the results on full-scale
raise the battery voltage is not possible unless special precau- cells are reported.
tions are taken to reduce leakage currents between cells. For
this application, we decided to use only parallel-connected
cells and a DC/DC converter to raise the voltage to a useful
level. This concept was also used in the unmanned under- 3. Experimental
water vehicle AUV-Demo in 1993 [1].
The CLIPPER battery has the shape of six cells oriented The test set-up consisted of the cell, a programmable load
around a central pressure hull as shown in Figs. 1 and 2. A and measuring circuits for cell voltage, current, temperature,
secondary battery is located inside the pressure hull. Three pressure loss and flow. Because the six cells are symmet-
dc to dc converters are located between the cells. They con- rically arranged, we only needed to test one cell to verify
vert the nominal one volt from the seawater cell to nomi- the battery performance in the full-scale test. The test was
nal 48 V for the CLIPPER system voltage. Each converter performed in a stationary test facility southwest of Bergen,
is connected to two cells. The secondary battery is charged Norway.
from the converters. Seawater was taken from the ocean into an axial pump,
The seawater cell is prismatic and contains 6 × 39 = then through a flow straightener to reduce vortices, then
234 parallel-connected magnesium rods and 5 × 38 = 190 through a venturi and into the cell. The aft end of the cell was
parallel-connected carbon fiber brush cathodes with metal open to the ocean. The flow was controlled by a brush-less
core. Cathode length was 350 mm and cathode diameter was electric motor in an oil-filled housing connected to an ax-
30 mm. Anode diameter was 22 mm at start of discharge. ial pump. Flow was measured using the differential pressure
Fig. 3 shows a top view of the inlet of one cell. Seawater over the venturi. The differential pressure over the cell was
enters the cell in one end and leaves in the other. On its also measured. Calibration was performed in a freshwater
way through the cell, the concentration of oxygen is reduced tank at FFI using ultrasonic flow measurement and water
and the concentration of the products of the cell reaction in- filled tubes for pressure measurement. One reference elec-
creased. These changes in the seawater chemistry increase trode (Ag/AgCl) was mounted in the compartment ahead of
with the length of the cell and decrease with increasing flow the cell and one in the cell outlet. Thus IR-loss in the elec-
velocity. Thus the cell voltage should increase with increas- trolyte in the cell is mainly included in the cathode potential
ing flow, as indeed observed, but as also the hydrodynamic measurements.
work (i.e. product of flow and differential pressure over the The equipment was mounted in an aluminium framework
cell) increases with increasing flow, there is an optimum flow and deployed at the seabed at a depth of 25 m. A cable was
through the cell where net power is at maximum. We also used for data transfer (RS485) and power (ac, 230 V) from
observed that there exists a minimum flow. Below that flow, shore. Advantech ADAM-modules were used for measure-
the cell clogs with the reaction products. Most experiments ment and control together with home made electronics. The
were made in the seawater laboratory on shore [1] with cells control programs were written using Advantech Genie ver-
with reduced electrode length (100 mm) and only two rows sion 3.0 and higher. This equipment allowed discharge at
of cathodes and three rows of anodes. In these experiments constant current or discharge at constant power.
parameters such as flow velocity, size of electrodes, distance Remote data-acquisition (using PC-Anywhere) made it
cell to cell and numbers of electrodes in the flow direction possible to evaluate the experiment from FFI.
were varied and cell performance and hydrodynamic power Logging and measuring equipment had a 110 Ah/12 V
battery back-up, sufficient for 48 h in case of power failure.
At the seabed, a lithium battery supplied local emergency
power.
On each cell, the following experiments were always per-
formed and in the following order:

• Pressure loss over the cell as function of flow (at open

circuit);
• Load curve: Cell voltage versus cell current
• Cell discharge curve: cell voltage versus time at 133 W
constant cell power.

Occasionally, the discharge was stopped and a new load

curve or pressure loss curve recorded followed by continued
discharge.
Fig. 3. Top view of the inlet of one cell. There are 6 × 39 = 234 anodes The anodes were individually weighted before and after
and 5 × 38 = 190 cathodes in each cell. each discharge.
 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239 235

0.2
1.5
0.4

1.4 0.6

E jj EA1 jj
0.8
EC1 jj
1.3
1

1.2
1.2

1.4

1.1
0 20 40 60 80 100 120 140 160 180 1.6
0 20 40 60 80 100 120 140 160 180
I jj I jj

Fig. 4. Cell voltage E [V] vs. cell current I [A]. Fresh cell (start of Fig. 5. Anode potential EA and cathode potential EC [V] vs. cell current
discharge). Test 2. I [A]. Fresh cell. Reference electrode located at the cell outlet. Test 2.

4. Results
is given as cm water column height. (One cm equals 100 Pa).
In test 2, the cell delivered 133 W for 504 h (67,032 Wh).
Three full cell discharge experiments were undertaken.
Charge delivered was 56,840 Ah.
The first was basically an equipment break-in test, while test
Test 3 was similar to test 2 with two important differences:
two and three were tests for battery characteristics. Fig. 4
shows the cell voltage and Fig. 5 the electrode potentials as 1. Initially, the cell was unintentionally exposed to an in-
function of the cell current. Flow was 8.5 l/s. Fig. 6 shows ternal short for nearly one hour, it was then taken back
the pressure loss over the cell and Fig. 7 the hydrodynamic to base, repaired and left at the quay for a week before
power lost in the cell as a function of flow. The pressure loss redeployment.

40

30

∆ P jj 20

10

0
0 2 4 6 8 10 12
Flow jj

Fig. 6. Pressure loss P [cm water] over the cell as function of flow [l/s]. Open circuit. Test 2.
236 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239

50

40

30

PH

20

10

0
0 2 4 6 8 10 12
Flow

Fig. 7. Hydrodynamic power lost PH [W] in the cell as function of flow [l/s]. Open circuit. Test 2.

2. The flow was reduced from 8.5 to 7.5 l/s. The results from during the test. The propeller rotation rate was constant dur-
test 2 and test 3 were fairly similar, therefore only data ing discharge, resulting in an increase in flow as the an-
from one of the tests are shown. odes were consumed and the gap between the electrodes
increased. The oscillations shown on the curves show the
Fig. 8 shows the cell voltage and cell current versus time. tidal frequency and are probably caused by variation in the
The cell power was a constant 133 W, thus cell current in- local water velocity (sea current). Fig. 10 shows the anode
creases as cell voltage decreases. The discharge was termi- and the cathode potentials versus time. The two reference
nated when the cell voltage fell to 0.6 V. Fig. 9 shows the electrodes gave fairly similar readings, only during the last
simultaneous pressure-loss over the cell, the corresponding part of the discharge was a small shift observed, indicating
hydrodynamic loss of power and the flow through the cell lower current density in the aft part of the cell.

200

150

100 E jj
100
I jj

50

0
0 50 100 150 200 250 300 350 400 450
Time jj

Fig. 8. 100 times the cell voltage E [V×100] and cell current I [A] vs. time [h] at 133 W load. Test 3.
 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239 237

Fig. 9. Upper curve: Pressure loss in the cell P batt [cm water], middle curve: Hydrodynamic power loss P hyd [W], lower curve: seawater flow [l/s]
vs. time [h]. Test 3.

0.5
EA1 jj

EA2 jj

EC1 jj

EC2 jj
1

1.5
0 50 100 150 200 250 300 350 400 450
Time jj

Fig. 10. Cathode potentials EC1 and EC2 and anode potentials EA1 and EA2 [V] vs. time [h]. Reference electrode no. 2 is close to the cell inlet,
reference electrode no. 1 close to the cell outlet. Cell load is 133 W. Test 3.

In test 3, we achieved 429 h at 133 W (57,057 Wh). Charge tential measurements. In test 2, the decreased consumption
delivered was 54,240 Ah. in the aft part of the cell was not observed.
In this test, initial anode weight was 56.2 kg. During dis-
charge, 41.2 kg was consumed. Fig. 11 shows the relative
anode consumption as a function of anode position. Anodes 5. Discussion
that were surrounded by 4 cathodes show the largest con-
sumption, whereas anodes at the corners of the cell show the Based on former experiences, an efficiency of 85% can
least. The figure shows a decrease in consumption toward be assumed for the dc to dc converter. Similarly, motor
the end of the cell in test 3. This correlates well with the po- efficiency is >90%. The result of the computational fluid
238 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239

100,0

90,0

80,0

70,0
Row 1
60,0 Row 2
Row 3
50,0
Row 4
40,0 Row 5
Row 6
30,0

20,0

10,0

0,0
0 5 10 15 20 25 30 35 40
Anode column number

Fig. 11. Relative anode consumption in wt.% as function of anode position. Test 3.

dynamics (CFD) was an expected electric power consump- The cost is also significantly lower. The negative side, be-
tion of 406 W for the propeller/pump combination given sides the low power rating, is the fact that there are locations
a vehicle speed of 2 m/s and a flow through the battery of where seawater batteries cannot be used, as the oxygen con-
53 l/s (8.8 l/s per cell). centration is too low. During one discharge of 400 h with a
The CLIPPER battery is composed of six seawater cells. flow through the battery of 45 l/s, approximately 65 000 t of
Thus electric power available for CLIPPER is 6 × 133 W × water goes through the vehicle. To avoid contamination of
0.85 = 678 W. Given that propulsion and pumping of water the cathodes, one should also avoid highly polluted areas or
through the battery requires 406 W, this leaves an average areas where the algae concentrations is high (low visibility
of 272 W for vehicle control and hotel loads. In addition, water). These limitations on applicability can be removed if
much higher power may be delivered for short periods as an acid solution of hydrogen peroxide is added to the elec-
the CLIPPER AUV will have a 12 cell serially connected trolyte [8], but at the expense of energy density and system
lithium ion battery of at least 120 Ah (ca 5.2 kWh). System complexity.
energy based on experiment 3 is 291 kWh plus the energy
stored in the buffer battery.
Our results show that the flow may be decreased to 45 l/s 6. Conclusion
with only marginal effect on performance, thus increasing
the allowable hotel load somewhat. This is important be- It has been shown that it is possible to make a vehicle,
cause pumping power is expensive (must be divided by pump using naturally occurring oxygen dissolved in seawater as
efficiency and by converter efficiency). oxidant and magnesium as fuel, which has the energy and
Comparing seawater batteries and conventional batteries power necessary to cross from Svalbard to North America
for AUV applications is difficult on a quantitative basis as via the North Pole or to patrol in an area for nearly 3 weeks.
the vehicles are very different. Seawater batteries operate At a nominal load of 133 W per cell, an endurance of
outside the pressure hull whereas conventional batteries 504 h has been achieved at a hydrodynamic loss of 24 W
usually are inside. Thus vehicle depth rating enters the and 430 h a hydrodynamic loss of 17 W. This performance
calculations. One way would be to use the total displace- is sufficient for a vehicle speed of 2 m/s (ca. 4 knots) with
ment, dry and wet volume for hotel load and hotel power a hotel load of approximately 270 W. These experimental
consumption and compare neutrally buoyant vehicles with results show that a range capability in excess of 1600 nautical
identical speed. Doing that, one finds that the energy of miles is well within reach. This long range, as well as the
the seawater battery system is very high and compara- large internal volume of 740 l of the pressure hull, makes
ble to what is possible with dense packing of primary the CLIPPER underwater vehicle a useful tool for civilian
lithium batteries in carbon fiber epoxy pressure containers as well as military oceanographic research.
of moderate depth rating. In contrast to primary lithium Compared to other power sources, the energy available
batteries however, the seawater battery does not increase for underwater vehicles using the seawater semi-fuel cell
the vulnerability of the mother vessel and it contains no is a factor of 3 better than the best available rechargeable
toxic ingredients that may do harm, even in a case of a batteries and comparable to primary lithium batteries. In
moderate fire. contrast to other high-energy power sources however, the
 Ø. Hasvold et al. / Journal of Power Sources 136 (2004) 232–239 239

magnesium dissolved oxygen seawater cell is intrinsically tonomous Underwater Vehicles, San Antonio, TX, 20–21 June 2002,
safe. pp. 89–94.
[4] Ø. Hasvold, N. Størkersen, Electrochemical power sources for un-
manned underwater vehicles used in deep sea survey operations, J.
Power Sources 96 (2001) 252–258.
References [5] Hasvold, Øistein, Seawater batteries for long term applications, in: T.
Keily, B.W. Baxter (Eds.), Power Sources, vol. 13, 1991, pp. 307–
[1] Ø. Hasvold, A magnesium-seawater power source for autonomous 317.
underwater vehicles, in: A. Attewell, T. Keily (Eds.), Power Sources, [6] Ø. Hasvold, H. Henriksen, B. Syversen, in: A. Attewell, T. Keily
vol. 14, 1993, pp. 243–255. (Eds.), Improvements in the rate capability of the magnesium-dissolved
[2] Ø. Hasvold, K.H. Johansen, O. Mollestad, N. Størkersen, The alka- oxygen seawater cell, Power Sources, vol. 15, 1995, pp. 149–162.
line aluminium/hydrogen peroxide power source in the HUGIN II [7] Ø. Hasvold, H. Henriksen, E. Melvær, G. Citi, B.Ø. Johansen, K.
unmanned underwater vehicle, J. Power Sources 80 (1999) 254–260. Kjønigsen, R. Galetti, J. Power Sources 65 (1997) 253–261.
[3] Ø. Hasvold, K.H. Johansen, K. Vestgaard, The Alkaline aluminium [8] M.G. Madeiros, R.R. Bessette, C.M. Deschenes, D.W. Atwater, Op-
hydrogen peroxide semi-fuel cell for the Hugin 3000 autonomous timization of the magnesium-solution phase catholyte semi-fuel cell
underwater vehicle, in: Proceedings of the 2002 Workshop on Au- for long duration testing, J. Power Sources 96 (2001) 236–239.