Paper presented by F.C. Walsh at UK Corrosion 2002, 22–24 October, Cardiff, UK.
ELECTROCHEMICAL AND PHYSICAL TECHNIQUES IN SUPPORT OF
THE CONSERVATION OF HISTORIC VESSELS IN THE SOLENT
K. Patterson, B.D. Barker
Applied Electrochemistry Group, University of Portsmouth, Portsmouth, PO1 2DT, UK.
F. C. Walsh (F.C.Walsh@bath.ac.uk)
Electrochemical Engineering Group, Department of Chemical Engineering, University of Bath,
Claverton Down, Bath BA2 7AY, UK.
This paper describes laboratory work which has been carried out in support of the conservation
of two vessels in the Solent area (Portsmouth, UK). The Holland 1 was the British Royal Navy’s
first submarine. Built in 1901, she sank off the Plymouth coast in 1913, and was raised in 1981.
Holland 1 is currently on display at the Gosport Royal Naval Submarine Museum. The removal
of chloride ions from the rust layers of the mild steel hull by alkaline washing are discussed. The
effect of stirring the sodium carbonate electrolyte on the rate of chloride ion removal from rust
films is examined. The M33 is a First World War ship, built in 1915. She stayed afloat for over
80 years until placed into a dry dock at the Portsmouth Historic Dockyard in 1997. This vessel is
being conserved using the electrolytic reduction method and the effect that this has on the
composition of the rust layers on the vessel are examined by X-ray diffraction.
Keywords: chloride ion removal, conservation of metals, electrolytic reduction, electrophoretic
migration, ion selective electrodes, iron oxides, marine corrosion, maritime heritage.
Once an artefact has been retrieved from a marine environment, the first aim of conservation is
to halt the current degradation processes and to stabilise the object so that further decay will be
minimised . It is usually considered important that the shape of the original artefact is
maintained. In the case of iron or steel artefacts the priority is the removal of aggressive ions,
such as chloride ions, to stop the present deterioration from continuing and also prevent
corrosion from recurring . This has been achieved by a number of different methods including
aqueous washing, electrolytic reduction, alkaline sodium sulphite reduction and lithium
hydroxide/alcohol washing. It is important that any scientific or historical data that the artefact
may provide is not irrevocably lost during the stabilisation and restoration processes.
The final aim of conservation is usually to produce a product that can be displayed to the public.
In most cases, items are displayed in a controlled environment i.e., low temperature and humidity
levels to minimise further degradation. When this is not possible, for example, with large
artefacts or vessels that are displayed in an open atmosphere, the application of a coating that is
preferably removable if needs be, is used to help prevent attack from the prevailing elements.
The present studies are part of a programme of laboratory work in support of the conservation of
a number of historic vessels using physico-chemical and electrochemical techniques. This paper
reports laboratory data relevant to the conservation two vessels in the Solent area, namely, the
Holland 1 submarine and the M33 monitor, bombardment ship.
The Holland 1 submarine has employed an alkaline washing method where the entire submarine
is soaked in a sodium carbonate solution, which was changed occasionally. The progress of
chloride ion removal from the surface of the vessel was monitored by taking regular samples of
the tank solution, which was analysed by potentiometric titration with silver nitrate solution. In
this research an electrochemical cell was designed using a chloride ion selective electrode that
could provide near instantaneous monitoring of the chloride content of a treatment tank. The
arrangement was used to study the effect that stirring the tank solution at increasing rates could
have on the rate of mass transfer of the chloride ions out of the artefact and into the wash
The M33 vessel has used the electrolytic reduction treatment method to assist the conservation of
the internal surfaces of steel plates near the bottom of the vessel. This method resulted in the rust
layers being reduced to magnetite (Fe3O4). As this iron oxide has a greater density (5.18 g m-3)
than others that may be present in marine iron and the volume of the film remains unchanged, the
result is a more porous structure . This facilitated the removal of chloride ions from the
artefact. A small-scale laboratory experiment was devised to examine the effect of electrolytic
reduction on the composition of the rust layers of an artefact. To achieve this, two hull samples
from the M33 have been subjected to electrolytic reduction and rust samples were taken
regularly. X-ray diffraction was used to determine the change in the magnetite content of the rust
layers during the experiment and results are summarised here.
2. THE HOLLAND 1: THE ROYAL NAVY’S FIRST SUBMARINE
2.1 Background of the Holland 1
Holland 1 (Figure 1) was the first submarine of the British Royal Navy and was launched in
October 1901 with a crew of nine. The hull was constructed from mild steel with a carbon
content of 0.1-0.15% and measured 19 m in length with a maximum diameter of 3.7 m. The
rivets on the hull were also mild steel but with a slightly higher carbon content of 0.18% . The
conning tower, torpedo tube and propeller were constructed from a two phase, / brass . The
vessel weighed around 105 tons and an electric motor powered by 60 lead acid batteries provided
enough power to enable her to dive to a depth of 30 metres and reach a top underwater speed of
seven knots. After 12 years of service, she was sold for scrap in 1913; whilst being towed to the
breakers yard, she sank just off the Plymouth coast.
2.2 Conservation of Holland 1
Holland 1 was raised in September 1981 and transported to a dry dock in Devonport where a
pressure wash was used to remove the concretion and marine life. The submarine was
transported to the Gosport Royal Navy Submarine Museum. To achieve this, the vessel had to be
cut into three sections which were welded back together upon arrival.
The exterior of the hull was coated with an anti-corrosion paint and the interior was treated with
a decorative paint. However, pre-treatment measures were not taken to remove the chloride ions,
and after standing for eleven years in a marine environment, the hull showed signs of severe
corrosion, particularly at any crevice sites, e.g., where the hull plates overlapped. The hull was
again cleaned with a pressure water wash and inspected.
The thickness of the hull had decreased from its original 11 mm to a range of 2.5-9.0 mm and
virtually all of the rivet heads were missing. It is reported that rust samples were removed and
analysed for chloride content by potentiometric titration with silver nitrate solution. The chloride
levels in surface films taken from the exterior of the hull were below the limit of detection with
this method (0.015 wt %) but the interior surface was found to have levels of 2.21% (wt/vol.). X-
ray diffraction on the rust samples showed that lepidocrocite (-FeO.OH), goethite (-FeO.OH),
haematite (Fe2O3), and silica were present on both the exterior and interior of the submarine;
magnetite (Fe3O4) was found on the interior and alumino-silicates on exterior rust samples .
As Holland 1 was too fragile to move again a fibreglass tank was constructed (on an existing
concrete base) around the submarine enabling her to be completely immersed in a wash solution.
Around 820 000 dm3 was required to fill the tank. Sodium carbonate was chosen for the wash
solution with the pH maintained at around pH 11. Sodium carbonate is an anodic inhibitor and
has the advantage of being relatively non-toxic. Due to the large volumes of solution required,
the tank solution was only changed occasionally but was re-circulated constantly. The progress
of the treatment was monitored by measurement of the chloride levels of the tank solution by
potentiometric titration. Rust samples were also taken occasionally for analysis when the tank
was being drained. Analysis of recent samples taken from various places inside the vessel
showed the chloride levels in the rust layers to range between 0.39 to 0.78 wt %.
This treatment and monitoring procedure continued until spring 2001 when it was decided that
the levels of chloride present in the rust layers of the vessel had been reduced to a sufficiently
low level (7 ppm). A tank constructed around the vessel will have controlled temperature and
humidity . Viewing access is provided to the public as well as an atmosphere in which the
vessel will suffer minimum further degradation.
2.3 Monitoring of the Chloride Release Rate using Ion Selective Electrodes:
The electrolytic cell is shown in Figure 2. Artificial artefacts were created by placing mild steel
samples, measuring 5 x 10 cm, in a Gallenkamp SSC450/E salt spray cabinet. They were
subjected to a continuous spray of sodium chloride solution 5% (wt/vol.) until covered with a
thick rust layer. The solution used in the bath was sodium carbonate at pH 9.5-10.0. The bath
solution was pumped through the cell at a volumetric flow rate of 0.38 cm3 s-1, until the reading
had stabilised. The chart recorder was started as the pre-rusted mild steel sample was placed into
the bath. The I.S.E potential was measured constantly and hence the chloride concentration
change over time in the bath was monitored on the chart recorder.
The experiment was first performed in a static bath with no stirring, then repeated with the bath
being stirred at three different rates of 123, 232 and 340 rpm which were measured using a RS
digital photo tachometer model RS163. The bath was stirred using a 4 cm long cylindrical
shaped magnetic flea and stirrer plate, to establish the effect that stirring may have on the mass
transfer of chloride ions from rust films. The experiments were performed in duplicate.
2.4 Chloride Ion Removal: Results and Discussion
Figure 3 shows data obtained when the I.S.E potential was recorded over time as a rusted mild
steel sample was placed in a tank of sodium carbonate solution at pH of 9.5-10.0. By calibration
of the electrodes the chloride content of the solution can be determined for a given potential.
Figure 4 shows a plot of the corresponding chloride ion concentration versus time.
The graphs show that the developed cell provided effective monitoring of the chloride content of
the bath. When the rusted mild steel sample was placed into an unstirred electrolyte, there was
an initial sharp increase in the chloride content of the bath which after 100 seconds began to
stabilise. In this case the transport of chloride ions from the rust layer and into the solution would
primarily be due to diffusion  with some natural convection.
When the bath was stirred, at three increasing rates of 123, 232 and 340 rpm, there was a
noticeable change in the rate of chloride released from the rust layers. Here movement of the
chloride species out of the rust layers could also occur by forced convection.
From the graphs it would appear that stirring caused an increase in the amount of chloride
removed from the rust layers. Although the samples used for the experiment were produced in
the same manner, the chloride content of the rust layers could vary considerably and so
conclusions about the concentration of chloride released from the rust layers are difficult.
During batch electrolysis of a constant area of chloride containing oxide, using a fixed volume,
V of electrolyte, and a fixed fluid flow conditions, the rate of chloride ion removal is often mass
transport controlled. The rate of removal is then limited by the rate of chloride ion movement
away from the steel surface. Under these conditions, the concentration of chloride ions in the
electrolyte will increase exponentially with time, giving rise to first order batch kinetics and the
potential of the chloride ion selective electrode (vs. a calomel reference electrode) will vary with
where: Eo = potential/mV at t = 0, E = potential/mV at a later time t, t = time/s and k = first order
rate constant/s-1. The equation may be written in logarithmic form as:
The effect of increasing the stirring rate of the bath on the rate of mass transport of chloride ions
from the rust layers and into the bulk solution was determined by plotting ln (E/Eo) versus time
for the initial linear region of Figure 3. An example of the plot obtained is shown in Figure 5 for
the data in Figure 3. For a first order reaction the slope of the graph would give a value for the
rate constant of the reaction according to equations (1) and (2) and the values are shown in Table
The apparent, first order rate constant, k depends on the flow conditions. k can be related to the
stirrer rotation rate, by:
Where x is the velocity exponent which can range from a value of 0.3 to 1 . is the stirring
rate in radians s-1 and is calculated as:
= 2f (4)
where f is the frequency in rev s-1 or (rev min-1/60).
Stirring speeds of 123, 232 and 340 rpm can be expressed as 12.9, 24.3 and 35.6 rad s-1,
respectively. A semilogarithmic plot of the rate constant (k) versus stirring rate () the slope of
which provided a value of x = 0.3 in equation (3). This value has been found to be very
dependent on the nature of the surface film as well as the flow conditions.
Transport of species (in this case chloride ions) can primarily occur in three ways namely
diffusion, convection and migration. As mentioned previously, in the unstirred bath the main
mode of transport for the chloride ions through the rust layers and into the bulk solution would
be by diffusion. Movement is driven by the differences in concentration and occurs down the
concentration gradient . When the artefact is submerged into the bath, the wash solution
would diffuse into the rust layers to form a concentrated salt solution, which could then diffuse
out of the rust layers and into the bulk solution. It can be seen both in Figures 3 and 4 that there
is an initial rapid linear increase in the chloride concentration of the bath. After 100 s of
immersion time, the concentration levels reach a plateau and this effect is said to occur as the
levels of chloride in the bath reach those in the rust layers of the artefact. North  has reported
that the rate-determining step in this process is the diffusion of chloride ions through the rust
layers. During the initial linear region of the graph the concentration of chloride ions released
from the rust layers into the bulk solution (c) can be related to the square root of the diffusion
coefficient for chloride ions (D) and the treatment time (t) by: .
It has been reported that values for the apparent diffusion coefficient of chloride ions through
rust layers from marine artefacts can range from 10-6 to 10-9 cm2 s-1 .
When the electrolyte is stirred, the movement of the chloride ions into the bulk solution could
now occur by both diffusion and convection.
The relationship between the mass transport coefficient and velocity was shown in equation (3).
The mass transfer coefficient (km) can be related to the Nernst diffusion layer thickness (N) and
the diffusion coefficient for chloride ions (D) by
As velocity increases, so does the value for the mass transfer coefficient. This results in a
decrease in the thickness of the Nernst diffusion layer. It can be said that by increasing the
stirring rate of the bath, the Nernst diffusion layer is decreased, and so the rate of mass transport
for chloride ions out of the rust layers of the artefact increases.
The results obtained in this research showed that in the unstirred bath the rate of chloride ion
removal was measured as 1.7 s-1 (see Table 1). Stirring the bath at 123 rpm caused the rate of
chloride ion removal to increase to 3.5. An increase in the stirring rate, from 123 to 232 rpm,
caused a further increase in the reaction rate to 5.4 s-1. When the stirring rate was increased
further to 340 rpm, however, the chloride ion removal rate showed little change at 5.5 s-1. This
shows how there is a limitation to the effect that the stirring rate can have on the removal rate of
chloride ions from the rust layers.
The ln-ln plots of k versus gave a relationship of k 0.3. Both the composition and the
morphology of the rust films can exert a large effect on the diffusion rate of the chloride ions
through the rust layers. We have experienced velocity exponent values of 0.5 in other trials. It
has been found in previous experiments that as the thickness of the rust layer increases the time
required to remove the chloride ions from the artefact can also increase . The composition of
the rust layer is an important factor in chloride removal as it can alter the porosity of the rust
layer. An increase in the magnetite content will cause the porosity to increase, which is thought
to lead to an increase in the diffusion coefficient for chloride ions . This would increase the
transfer rate of chloride ions through the lattice. The speciation of chloride in the solid surface
oxide layers is clearly an important aspect, which can control the rate and ease of chloride ion
dissolution into an aqueous electrolyte phase.
3. M33: A FIRST WORLD WAR MONITOR SHIP
3.1 Background to the M33 Vessel
Following the outbreak of the First World War in 1914, Winston Churchill (First Lord of the
Admiralty) and J. Fisher (First Sea Lord) initiated a program of ship construction. This included
a new design of five small six-inch gun monitors, which were vessels with a shallow draft (of
just 6 ft) that would enable them to sail into the very shallow waters off coastlines. The M33
vessel is shown in Figure 6.
Along with the other four monitor vessels, M33 was built at Harland and Wolff, in Belfast, in
just three months. She had a displacement of 580 tons, an overall length of 54.0 m and a
maximum width of 9.5 m. An error in the design of the bow meant that she did not sit
sufficiently low in the water. When this problem was coupled with an undersized rudder, it
caused steering problems. 4 tons of permanent ballast (concrete) was required to rectify the
problem. After sea trials, where a top speed of 9.6 knots was reached, she was launched on the
22nd May 1915 .
On 26th June 1915, she departed for the Dardanells where her first assignment was to support the
landings made by the British, French, Australian and New Zealand forces on the Gallipoli
Peninsula. Until November 1915, her duties also included patrolling the coast. Despite being in
the midst of continual firing, she was never hit. On the 11th January 1916, she left for her next
assignment which involved guardship duties in Salonika and again, even though a number of
bombs were dropped all around her, this lucky ship still received no hits. In 1925, she received
the name H.M.S. Minerva; she is now known by her original name, M33.
Since the war she has had several different roles including being converted into a minelayer and
a floating workshop. In 1991, she was sold to Hampshire County Council and berthed in the
Naval Base at Portsmouth.
3.2 Conservation of the M33
On 23rd April 1997, M33 was placed into Number 1 dry dock at the Portsmouth Historic
Dockyard which had been prepared for the vessel with the addition of 6 ft high wooden support
blocks. The hull of the vessel that had been afloat for more than 80 years could finally be seen. A
high-pressure water jet wash was employed to remove the worst of the various marine life that
had encrusted the hull. Samples of rust were analysed for chloride content by potentiometric
titration with silver nitrate solution (0.1 mol dm-3), values of 2.6-3.4% (wt/vol.) were found for
the exterior of the hull and 0.76-1.14% (wt/vol.)  for the interior of the hull. The corrosion
products were analysed by X-ray diffraction. Both the interior and exterior samples analysed
were found to be composed mainly of goethite (-FeO.OH) and lepidocrocite (-FeO.OH).
Analysis of the hull showed that it consisted of approximately 80% ferrite (-Fe) and 20%
pearlite (layers of Fe3C and -Fe). The carbon content of the steel was calculated to be 0.17%
The electrolytic reduction method was chosen to assist in the conservation of the internal hull
surfaces of M33 in the period 1998-2001. Each section of the hull was taken in turn and is made
cathodic with respect to 316 stainless steel expanded mesh anodes in a 0.6 g dm-3 sodium
carbonate electrolyte, typically at 0.35 A m-2 current density on the hull surface and 18 A m-2 on
the anodes. After electrolysis, each hull section was left to soak in a solution of sodium carbonate
with a pH above 9.5.
The conservation treatment would, ideally, be followed by the vessel being placed into a
controlled environment but due to the size and location of M33 it means that this may not be
practical and the ship will be constantly subjected to a marine environment. It is planned to apply
a protective organic coating to the interior and exterior of the hull.
3.3 X-Ray Diffraction Analysis of the Change in the Composition of Rust Layers on Hull
Samples Subjected to Electrolytic Reduction: Experimental Details
The apparatus was arranged as shown in Figure 7. The mild steel hull sample measuring
approximately 27.8 x 4.7 x 0.8 cm was immersed in approximately 3.2 dm3 of demineralised
water containing enough sodium carbonate to increase the pH above 9.5. The sample was then
connected to a power supply. A piece of platinised titanium measuring 20 x 5 cm was used as an
inert anode and was placed parallel to the hull sample at a distance of approximately 3 cm. A
voltmeter was connected between the sample and reference electrode, an ABB saturated calomel
electrode (S.C.E) was used. Enough current was supplied to maintain the potential of the sample
at or below –0.791 V vs. S.C.E., which would place the mild steel sample in the zone of
immunity. The cell potential was measured by connecting an Iso-Tech IDM digital voltmeter
between the sample and the inert anode.
Rust samples were removed from the hull piece prior to the experiment and weekly for five
weeks. The wet rust samples obtained were first dried in an oven at a temperature of 120C for at
least 4 hours after which they were crushed in a pestle and mortar to give a fine powder. This
was then further ground in ball mill for 10 minutes using a Pitchford mixer to give a
homogeneous mixture. This also ensured that all the samples were ground to the same degree.
The powdered rust samples were spread evenly in an aluminium tray and a diffraction pattern
produced for each sample at angles 2 between 5 and 60. The XRD instrument used was a
Phillips PW1710 with nickel-filtered, Cu K radiation. The working voltage was 40 kV and the
current was 20 mA.
For quantitative measurement of the change in levels of magnetite in the rust layers, a calibration
chart was constructed. Calibration standards of 0, 25, 50, 70 and 100 wt % were prepared by
accurately weighing out the required amount of magnetite (Fe3O4) into a small ball mill. The rest
of the powdered sample was made up using haematite (Fe2O3) that was also accurately weighed
into the mill and used due to its availability and to provide a similar background as would be
present in the samples. Both compounds were previously dried in an oven at 120C for at least
four hours. The two powders were then mixed in the same way as the samples i.e., for 10
minutes in the small ball mill with the Pitchford mixer. A diffraction pattern was produced for
each standard in the same way as for the samples from which a calibration chart was produced
via analysis of the intensities of the main magnetite peaks.
3.4 Results and Discussion: M33 Monitor Vessel
X-ray diffraction patterns were obtained for rust samples taken before the experiment began and
then weekly over the next five weeks.
The diffractogram for an unelectrolysed sample showed that the rust layer initially consisted of
magnetite (Fe3O4), goethite (-FeO.OH), lepidocrocite (-FeO.OH), akaganéite (-FeO.OH), this
agrees with the literature that states that these are the four predominant iron oxides present in
marine rust .
The overlaid patterns showed that, after 5 weeks of electrolytic reduction, there was a
considerable increase in intensity of the magnetite peaks both in value and in relation to the
peaks for the other oxides present. For example, the peak at approximately 35 degrees 2
increased in intensity from a value of 80 to 180 in the first experiment and from approximately
160 to 210 in the repeat. At the end of the experiment the peaks for lepidocrocite and akaganéite
virtually disappeared and so it could be said that these oxides were no longer present above
detectable limits. The three main goethite peaks were all slightly reduced in the first experiment
but show little change in the repeat.
The results for the quantitative measurement of the average magnetite content of the rust layers
over treatment time are shown in Table 2. The results show that in the first experiment the
magnetite content increased from an average of 10.2% to 39.4% after five weeks and in the
repeat the increase was from an average of 27.4% to 43.8%. However, the increases observed
were initially insignificant until the third or fourth week of the experiment. After this time a
dramatic increase was observed.
Previous work on atmospherically produced rust layers also found that lepidocrocite was reduced
to magnetite by electrolytic reduction [15,16] these experiments were carried out at slightly
acidic to neutral pH solutions. Kuch  suggested that the reaction occurring might be:
3-FeO.OH + H+ + e- Fe3O4 + 2H2O E0 = 1.73 V (vs. S.H.E) (7)
However, the research in  was carried out in aqueous solutions at pH>9.5 so reaction (8)
shows the reduction reaction for lepidocrocite that may occur.
3-FeO.OH + O2 + H2O + 3e- 5OH- + Fe3O4 E0 = 1.63 V (vs. S.H.E) (8)
Both of these reactions assume that the conversion is a direct process and does not involve the
formation of any intermediate products. Transformations in rust can occur in two ways, the first
is topotactically where the reaction takes place in the solid phase and involves the internal
rearrangement of atoms. The other method of transformation is by reconstruction where the rust
phase dissolves and re-precipitates from solution as a new phase . The reactions in this
research are most likely topotactical. Lepidocrocite has a cubic close packing (ABC) of O2- and
OH-. The O2- of magnetite also forms a cubic close packed (ABC) structure. This similar packing
of anions means that the conversion of lepidocrocite to magnetite could occur readily.
Work carried out on atmospheric rust layers did not cover the reduction of akaganéite, as this
type of rust is only formed in chloride containing environments, such as seawater. It is suggested
that chloride ions are incorporated into the lattice and appear to play a role in the stability of
akaganéite . Akaganéite has been previously found to be reduced to magnetite in alkaline
solutions (pH 7-12) at 25C in the presence of Fe2+ ions . The Pourbaix diagram for iron in
water is shown in Figure 8, from which it can be seen that in the approximate range of pH 9 to
pH 13, magnetite (Fe3O4) would be the most thermodynamically stable species. In this research
the experiments were carried out at a pH of 9.5-10.0. When electrolytic reduction is carried out
at an elevated pH, the formation of magnetite would be favoured. For the electrolytic reduction
of akaganéite, the reaction could follow that shown by reaction (8).
In agreement with previous work [15-17,20-21], goethite was not reduced. This may be due to
the the hexagonal close packing arrangement of anions in goethite which means that reduction to
magnetite would involve the transformation to a cubic close packed structure . Previous
investigations into the topotactic transformation of Fe(OH)2 (which also has a hexagonal close
packed (AB) arrangement of O2- and OH-) to magnetite (ABC) found that transformation
occurred via intermediates green rust I followed by green rust II. This was thought to occur
because these phases contain both hexagonal and cubic close packed layers of anions. With
green rust I the stacking is ABCBCACAB and for green rust II is ABAC. The formation of these
intermediates allows the proportion of cubic close packed anions in the lattice to be slowly
increased. In this way a hexagonal close packing of anions could be slowly converted to a cubic
The intermediate compounds green rusts I and II were not identified by X-ray diffraction in this
research. If they were not formed, then this may be why goethite was not reduced to magnetite. It
has been found that chloride ions are required for the formation of green rust I and are
incorporated into the structure . For the formation of green rust II the presence of sulphate
ions is required. This would have been absent from the electrolyte used in the present studies.
Experiments on the effect of stirring on the mass transport of chloride ions out of rust layers, into
a carbonate washing solution, showed that stirring the treatment bath can increase the rate at
which the ions are removed from the rust layers of an artefact. As aqueous/alkaline washing
methods have been found in the past to progress at a very slow rate this small modification may
help to shorten treatment times. The results did show that there was a limit to the effect that
stirring the bath can have and so a gentle stirring rate of between 100 and 200 rpm is
recommended. The electrochemical cell which was designed for the experiments proved to give
near instantaneous monitoring of the chloride levels of the treatment bath. The current method of
monitoring the progress of a chloride removal treatment is to remove a sample from the bath and
analyse by potentiometric titration. This can be laborious and not always practical. The cell set
up used in this research provides an alternative method of chloride determination and has the
advantage in that it can be easily employed to provide constant, and on-site, information as to the
progress of chloride removal from artefacts.
The effect of electrolytic reduction on the composition of rust layers on iron or steel artefacts
was investigated. The results showed that the marine rust analysed in this research predominantly
consists of magnetite (Fe3O4), goethite (-FeO.OH), lepidocrocite (-FeO.OH), akaganéite (-
FeO.OH). The quantitative X-ray diffraction analysis showed that the electrolytic reduction
method of conservation causes the magnetite content of the rust layer to increase considerably
after 4 or 5 weeks of treatment. This may cause an increase in the rate of chloride removal from
the rust layers.
The authors are grateful to Peter Lawton (Treadgold Musem, Bishop Street, Portsmouth PO1
3DA, UK) for access to the M33 vessel and to Commander K. Tall (Gosport Royal Naval
Submarine Museum) for access to the Holland 1 submarine. Thanks are due to Derick Weights
(University of Portsmouth) for his assistance with the X-ray diffraction analysis.
1. Favre-Quattropani L., Groening P., Ramseyer D., Schlabach L., The Protection of Metallic
Archaeological Objects using Plasma Polymer Coatings, Surface and Coatings Technology,
March 2000, 125 (1-3), 377-382.
2. Walker R., Stabilisation of Marine Iron Artefacts, British Corrosion Journal, 1996, 31 (1),
3. Kentish P.L., Dillon B.I.. Conservation, Preservation and Presentation of Iron and Steel
Artefacts. Conference on the Protection of the Engineering Heritage, 1982, pp. 46-49.
4. Barker B.D., Barnes S., Bevan T., Jones S., Walsh F.C., The Corrosion and Conservation of
Two Royal Navy Ships Built in the Early 1900s, UK Corrosion 95 Conference, Harrogate,
5. Photograph courtesy of Dr. B. D. Barker, University of Portsmouth.
6. Royal Navy Submarine Museum, Web Site,
7. Walsh F.C., A First Course in Electrochemical Engineering, The Electrochemical
Consultancy, Romsey, 1993.
8. Walsh F.C., The Kinetics of Electrode Reactions: Part II - Mass Transfer and Mixed Control,
Transactions for the Institute of Metal Finishing, 1992, 70 (2), pp. 95- 99.
9. North N.A., Pearson C., Washing Methods for Chloride Removal from Marine Iron
Artefacts, Studies in Conservation, 1978, 23, 174-186.
10. Pearson C., Conservation of Marine Archaeological Objects, Butterworths, Oxford, 1987.
11. Photograph courtesy of K. Patterson 1997.
12. Buxton I., M33 - A Short History. Museum Pamphlet, Produced by Tregonwell’s Museum,
Portsmouth, 1993; http://www.envf.port.ac.uk/olc/m33/conserv.htm
13. Patterson K., The Corrosion Problems on H.M.S. Minerva (M33). Final Year Undergraduate
Project, University of Portsmouth, 1997.
14. Cornell R.M., Schwertmann U., The Iron Oxides, Structures, Properties, Reactions,
Occurrences and Uses, VCH, Cambridge, 1996.
15. Suzuki I., Masuko N., Hisamatsu Y.. Electrochemical Properties of Iron Rust, Corrosion
Science, 1979, 19 pp. 521- 535.
16. Perdomo J.J., Chabica M.E., Song I., Chemical and Electrochemical Conditions on Steel
under Disbonded Coatings: the Effect of Previously Corroded Surfaces and Wet and Dry
Cycles, Corrosion Science, 2001, 43 (3), pp. 515- 532.
17. Kuch A., Investigations of the Reduction and Re-oxidation Kinetics of Iron (III) Oxide
Scales Formed in Waters, Corrosion Science, 1988, 28 (3), pp. 221-231.
18. Pourbaix M., Atlas of Electrochemical Equilibria in Aqueous Solutions, Pergamon Press,
19. Ishikawa T., Katoh R., Yasukawa A., Kandori K., Nakayama T., Yuse F., Influences of
Metal Ions on the Formation of -FeOOH Particles, Corrosion Science, 2001, 43 (9),
20. Kendall K., PhD Thesis, “Gaseous Reduction of Archaeological Ironwork”, Portsmouth
21. Archer J., PhD Thesis, “Conservation of Archaeological Artefacts by Thermal Methods",
Portsmouth Polytechnic, 1991.
22. Stratmann M., Bohnenkamp K., Engell H.J., An Electrochemical Study of Phase-Transitions
in Rust Layers, Corrosion Science, 1983, 23 (9), pp. 969- 985.
23. Misawa T., Hashimoto K., Shimodaira S., The Mechanism of Formation of Iron Oxide and
Oxyhydroxides in Aqueous Solutions at Room Temperature, Corrosion Science, 1974, 14
(14), pp. 131- 149.
24. Refait P.H., Abdelmoula M., Génin J.M.R., Mechanisms of Formation and Structures of
Green Rust One in Aqueous Corrosion of Iron in the Presence of Chloride Ions, Corrosion
Science, 1998, 40 (9), pp.1547-1560.
Apparent first order rate constant, k
from Figure 3
Table 1. Apparent first order rate constants calculated from the slope of the
linear region on the semilogarithmic plot of E/E0 vs. time in Figure 3.
Time of electrolytic
Table 2. Average magnetite content of samples vs. electrolysis time (using X-ray
Figure 1. Holland 1 at the Gosport Royal Navy Submarine Museum prior to conservation
in 1982 .
Figure 2. Arrangement for ion selective electrode monitoring of the release of
chloride ions from a rusted mild steel sheet (5 x 10 cm).
Figure 3. ISE potential vs. time for a rusted mild steel sample in a bath of sodium carbonate at
pH 9.5-10.0, with increasing stirring rates, at 22C. Magnetic stirrer bar speeds:
(a) 0 rpm, (b) 123 rpm, (c) 232 rpm, (d) 340 rpm.
Figure 4. Chloride ion concentration in the carbonate electrolyte vs. time,
determined from the data in Figure 3. Magnetic stirrer bar speeds:
(a) 0 rpm, (b) 123 rpm, (c) 232 rpm, (d) 340 rpm.
020 40 60 80 100 120 140 160 180 200
Time / s
E vs. Ag/AgCl/1M KNO
3 / mV
050 100 150 200
Time / s
Cl- concentration / ppm
Figure 5. Natural logarithm of (E/Eo) vs. time, taken from the data in Figure 3.
E = Chloride ISE potential at time t; Eo = Chloride ISE potential at time zero.
Magnetic stirrer bar speeds: (a) 0 rpm, (b) 123 rpm, (c) 232 rpm, (d) 340 rpm.
Figure 6. M33 after being placed into dry dock for conservation at the Royal Navy Base at
Portsmouth Historic Dockyard in April 1997 .
0 5 10 15 20 25
ln (E/Eo / mV)
Time / s
Figure 7. The experimental arrangement for electrolytic treatment of the
rusted hull plate samples to remove chloride ion.
Figure 8. Equilibrium potential versus pH (Pourbaix) diagram for iron in water at 25C.
(a) oxygen evolution line, (b) hydrogen evolution line .