Corrosion Science, 1972, Vol. 12~ pp. 227 to 246. Pergamon Press.

Printed in Great Britain

MECHANISM OF ATMOSPHERIC RUSTING *

U. R. EVANS and C. A. J. TAYLOR

Department of Metallurgy, Cambridge University, Pembroke Street, Cambridge, England

Abstract--The present paper describes a research on Atmospheric Rusting, which has been in progress
for 5 y in the Department of Metallurgy at Cambridge University. The discussion is designed to
present the connection between the electrochemical mechanism of atmospheric corrosion with that
of corrosion under immersed conditions.
R~sum&---Lepr6sent m6moire d6crit une recherche relative h la formation de rouille atmosph&ique,
qui a 6t6 poursuivie pendant cinq armies au d6partement de m6tallurgie de l'Universit6 de Cambridge.
La discussion a pour objet de presenter le lien entre le m6canisme 61ectrochimique de la corrosion
atmosph6rique et celui de la corrosion dans des conditions d'immersion.
Zusammenfassung--Die vorliegende Abhandlung beschreibt eine Untersuchung tiber das atmo-
sphfirische Rosten, die seit 5 Jahren in der Abteilung fiir Metallurgie an der Universit~.t Cambridge
durchgeftihrt wird. Die Erlfiuterungen dienen dem Zweck, den Zusammenhang zwischen den elektro-
chemischea Mechanismen der atmosph/irischen Korrosion und der Korrosion im untergetauchten
Zustand zu zeigen.

INTRODUCTION
Knowledge available in 1965
IT IS generally agreed that the mechanism of corrosion under immersed condition is
electrochemical. For plates of iron partially immersed in potassium chloride solution,
currents passing between cathodic and anodic areas have been measured and found
to be equivalent, in the sense of Faraday's law, to the corrosion-rate determined
gravimetrically. 1The reason for the rapidity of electrochemical corrosion is well known.
Direct chemical attack would produce a solid film which would bring attack almost to
a halt; electrochemical action often produces soluble matter at anodic and cathodic
areas, insoluble substances being thrown down at a distance from the surface, where
they fail to protect.
In corrosion by drops of salt solution placed on a horizontal iron surface, the
currents passing radially between the cathodic periphery where oxygen is readily
renewed, and the central anodic area which is less well aerated, have been elegantly
demonstrated by Blaha; ~ in a magnetic field, the liquid as a whole rotates; when the
magnetic polarity is reversed, the direction of the rotation reverses also.
Regarding atmospheric corrosion there is less agreement. Moist air, even if
saturated, produces little rusting on iron, provided that dust and pollution are
excluded; iron plates have been kept over water in a closed vessel for a month and
showed only tiny dots of rust at the end.3 The action of settled dust in causing rusting
of iron and steel has long been known, but Vernon 4 was the first to show that not aU
kinds of dust produce the effect; silica particles cause no rusting, whereas ammonium
sulphate is very effective. He showed also that rusting is only serious if the relative
humidity exceeds a certain value.
*Manuscript received I0 June 1971.
227
228 U.R. EVANSand C. A. J. TAYLOR

Most salt particles produce rusting if the humidity exceeds a critical value, which
varies with the salt; in the case of hygroscopic salts such as the chlorides of mag-
nesium, zinc or lithium, rusting can occur at very low humidities. Bukowiecki 5
showed that generally the critical humidity is close to the vapour pressure of a saturated
solution of the salt in question. This .means that rusting will only occur if the air is
sufficiently damp to ensure that a film of solution would not dry up; we may picture
--under the conditions where rusting is set up by dust--a tenuous film of saturated
salt solution between the particle of solid salt and the metal; a visible puddle of liquid
is not needed. Under these circumstances, the particle will hinder renewal of oxygen in
the centre but not at the periphery of the moisture-patch. If Blaha's demonstration of
radial currents within a rounded drop is accepted as evidence of electrochemical action,
conditions for such currents should be even more favourable below the salt particles.
There is no need to look further for an explanation of the rust-spots produced at points
where such particles settle.
However, one of the most potent causes of atmospheric corrosion in urban and
industrial districts is the sulphur dioxide produced by the burning of fuel. Here no
solid substance is present, but clearly if the sulphur dioxide dissolved in an adsorbed
film of water is oxidized to sulphuric acid, this will collect further water from the
atmosphere. The acid will attack the iron to give ferrous sulphate, which itself is
oxidized to produce ferric sulphate and hydrated ferric oxide or basic ferric sulphate;
the hydrolysis of ferric sulphate will furnish further hydrated ferric oxide or basic
sulphate. There is little doubt that the first traces of brown rust-like substance are
formed in that way.
The extensive work of Hudson s in several parts of the world has shown that
atmospheric corrosion is most rapid in places where the sulphur dioxide content of the
atmosphere is high (the correlation has been studied statistically), 7 whilst Schikorr s
has shown that, in a given locality, rusting is more rapid at seasons of the year when the
sulphur dioxide is highest. Schikorr has also established the fact that a small amount
of sulphur dioxide will produce a relatively large amount of rust; in one set of tests,
each sulphur dioxide molecule was found to lead to anything between 15 and 40
molecules of rust--according to the season. Evidently some cyclical change must be
involved, whereby a sulphur compound operates to produce rust, being itself
regenerated in the process, so that it can then produce further rust. Two possible
cycles have attracted interest:
One of these, the Acid Regeneration Cycle, was suggested in the early writings of
Vernon 9 and has been considered in detail by Schikorr and other workers. The presenta-
tion has varied, but the essential feature is the oxidation of ferrous sulphate to ferric
sulphate, which hydrolyzes to give hydrated ferric oxide (rust) and sulphuric acid;
that acid attacks iron to give fresh ferrous sulphate, which, by oxidation and hydrolysis,
produces yet more acid--and so on. It should be borne in mind that hydrolysis will not
necessarily liberate all the acid; some may become insoluble as basic ferric s u l p h a t e n
a point discussed later. That, however, is not fatal to the theory, and indeed may be
said to provide a suitable explanation of Schikorr's discovery that a finite amount
of sulphur dioxide produces a finite amount of rust--not an infinite amount, as would
be expected if the hydrolytic liberation of acid could proceed to completion.
The other proposed mechanism is the Electrochemical Cycle. So soon as even small
 Mechanism of atmospheric rusting 229

amounts of hydrated ferric oxide (rust) and ferrous sulphate have appeared on the
surface, conditions become favourable to electrochemical changes connected with the
cell Fe IFeSO4] FeOOH. At certain points the anodic reaction

Fe = Fe 2+ -k- 2e

will occur, whilst at other points the two electrons will be used up in the cathodic
reduction of the rust to magnetite

Fe 2+ ÷ 8FeOOH + 2e = 3Fe304 + 4H20.

The magnetite will quickly be re-converted by atmospheric oxygen to rust

3Fe304 + 0"7502 q- 4"5H20 = 9FeOOH,

but now there are nine molecules of FeOOH instead of only eight--a net gain of one,
due to the entry of one Fe "+ ion. The amount of Fe '+ used up is equal to that re-
generated by the anodic reaction. Since iron enters the liquid as an ion at one point
and leaves it at another point, the newly formed rust will not be protective and the
reaction can continue.
These changes were demonstrated experimentally 10 y ago. 1° The cell
FeIFeSO4 solutionIRust (supported on copper),
was found to yield an electrical current, the iron being the anode. The current, how-
ever, gradually died away, and it was then found that the brown colour of rust had
been replaced by the black colour of magnetite. If, however, the black magnetite was
exposed wet to air, it became brown again; the electrode could then be returned to the
cell as cathode, and would provide about the same amount of current as before.
Perhaps one reason for the non-acceptance of this mechanism--apart from the
delay always shown by scientists in accepting new ideas*--may be the belief that the
magnetite is a very stable substance. Doubtless a perfect magnetite crystal does resist
oxidation in damp air, but newly formed magnetite (produced, for instance, by the
action of alkali on a solution containing Fe 2+ and Fe s+) soon loses its magnetism and
changes its colour when exposed to water with access of air.
It was felt that an objective study of the subject, designed to discover which, if
either, of the mechanisms was adequate to explain the facts, was long overdue. That
was the reason for this new research.
Exploratory research published in 196571 Before starting measurements, a qualita-
tive reconnaissance seemed appropriate. Specimens were exposed in closed tubes to
air containing regulated amounts of moisture and sulphur dioxide. The first change
was the formation of a faint mist of moisture on the surface which could be wiped off
with filter paper and tested; it was found to contain ferrous iron. Later, dark spots
appeared, and spread gradually over the surface; the solid matter was first grey and
*The electrochemicalmechanism of corrosion by liquids was only accepted after a long period of
scepticism; but perhaps a better example of delay is that of Mendelian principle in biology, which
only achieved general acceptance 35 y after its discovery by Mendel.
230 U.R. EvANsand C. A. J. TAVtOR

then became brown; it contained ferrous sulphate and hydrated ferric oxide, the
proportion of the latter increasing as time continued. If at the time when the layer
contains both hydrated ferric oxide and ferrous sulphate, a specimen was transferred
to a new tube containing moist air but no sulphur dioxide, the formation of fresh rust
still continued.
These experiments show that a soluble compound (in this case ferrous sulphate)
is formed before any insoluble rust appears, and that sulphur dioxide is only needed
as the source of the ferrous sulphate; when once the latter has been formed, rusting
can continue even when there is no sulphur dioxide in the gas-phase. The continued
formation of rust from ferrous sulphate could be explained either by the Acid Regenera-
tion Cycle or by the Electrochemical Mechanism. In the first case, the presence of much
acid and soluble ferric sulphate on the surface would be expected, and the tests applied
did not support this; the amount of water-soluble material decreased as time proceeded
and only a trace of water-soluble ferric iron could be detected. Although not conclusive,
the results favoured the Electrochemical Mechanism.
In recent years, there has been greater readiness to consider an electrochemical
interpretation. An electrochemical view of atmospheric rusting has been put forward
by Schwarz, x2 Schikorr 13 has considered the appropriateness of re-expressing his
theory in electrochemical form, whilst Barton 14 has advocated the introduction of new
electrochemical concepts into the subject of atmospheric corrosion. Nearly all
authorities are agreed on one point--that ferrous sulphate plays an important part.
The distribution and composition of the ferrous sulphate in rust formed on iron
exposed out-of-doors has been the subject of much study; it is found 15,~8 to be
F e S O 4 . 4 H 2 0 - - n o t FeSO4.7H~O as was once supposed; its situation in a rust-coat
varies with time and doubtless with locality. 17,1s

New experiments
Materials. Experiments of Sets A to H were carried out on iron sheet containing
0.006 C, 0.001 Si, < 0.005 Mn, 0.0066 S, 0.0017 O, 0.0016N and 0-004 A1. For the
later sets, steel sheet containing 0.065 C, 0.34 Mn, 0.032 S, 0.015 P, 0.08 Cu and
0.08 Ni was used, but most experiments were repeated on the same steel electro-
plated with iron from a bath made from ferrous sulphate of A.R. quality, which con-
tained < 0.001 Cu, < 0.003 Pb, < 0.025 Mn and 0.01 Zn as chief metallic im-
purities. Set O was carried out on both steel and iron sheet.
Surface preparation. Specimens were abraded, being degreased with xylene and
acetone both before and after abrasion. In cases where freedom from inclusions of
soluble matter was important, the abrasive used was dry "Solihull" pumice powder
(the "coarse" variety), which had been freed from soluble salts by repeated boiling
with doubly distilled water, and was applied on unbleached linen, which had itself
been repeatedly boiled with doubly distilled water.
Rusting procedure. Experiments of Sets L, M, N and Q required specimens carrying
an adherent rust-layer. This was obtained by stroking the iron surface with the edge of
a rectangle of sheet rubber wetted with M/10 ferrous sulphate, leaving a shallowpuddle
o f solution on the area to be rusted which was allowed to dry up; a frosty layer of
microcrystals remained. The specimen was placed over water in a room thermostatted
at 25°C for one or more days; a single night produced a satisfactory rust-deposit.
 Mechanism of atmospheric rusting 231

Tube experiments. The number of factors capable of independent variation

demanded a large number of experiments; a simple procedure was therefore essential.
Experiments A to H were carried out in glass boiling-tubes, 18 cm long and 2.7 cm
internal dia. (Fig. l(a)). The specimens (r) were rectangles 3.0 × 1.6 cm, wedged into
cylinders of waxed paper (b), and supported by false bottoms (c). The cylinders were

~.b

(el
(d)
(b)
(a)

FIo. 1. Experimental vessel. The specimen(r) is wedged into the waxed-paper tube Co),
being supported by false bottom (c); the waxed-paper tube is itself placed in a glass
tube (a), being supported by false bottom (d). A small amount of dilute H~804, usually
containing KzS2Os,is placed at the bottom of (a) but does not wet (b).

introduced into the glass tubes, which contained at the bottom a small amount of
sulphuric acid of a concentration chosen to provide the desired humidity; (in most
cases, the appropriate amount of potassium metabisulphite was added to the acid, so
as to liberate sulphur dioxide). A second false bottom (d) prevented the waxed cylinder
from entering the liquid. After introduction of the cylinder, each tube was stoppered,
and, by means of a magnet, the specimen with its waxed cylinder was raised to the top
and allowed to fall; this procedure, designed to distribute water vapour and sulphur
dioxide throughout the gas phase, was repeated three times. The tubes were kept in a
room thermostatted at 25°C for the desired time.
In the one-stage experiments (Sets A and B), the specimen remained in the same
tube throughout the chosen experimental period; it was then removed, dried for 24 h
at 25°C over saturated magnesium chloride solution containing some sofid magnesium
chloride, and weighed with the rust adhering; finally it was derusted and reweighed.
Magnesium chloride (used by Schikorr) provided an atmosphere designed to remove
adherent and capillary water from FeOOH without dehydration to FeaO3.
The derusting was carried out with 2-5N hydrochloric acid in a cell (j) (Fig. 2)
containing a steel frame (f) of relatively large area--so as to provide for reductive
232 U.R. EVANSand C. A.J. TAYLOR

--f

~d

FIG. 2. Derustingvessel The glass jar (J), fitted with a large steel frame (f) is filled with
dilute HCI. The specimen (r) makes electrical contact with the frame--a plan which
prevents serious attack on the iron substrate of the specimen.

dissolution of the iron as ferrous chloride without attack on the iron basis of the
specimen. Reductive dissolution as ferrous chloride, which--as shown by Pryorl~--is
a much quicker reaction than direct dissolution as ferric chloride, involves anodic
attack on the metal basis by cathodic reduction of a ferric to a ferrous compound. If
the specimen is introduced into acid without precaution, the iron basis provides the
anode, and a serious error is introduced, which is not necessarily avoided by the use of
a "pickling inhibitor"--a point sometimes overlooked; by providing the steel frame
as a large anode, the error can be made acceptably small.
Experiments C - H were conducted in two or more stages. First the specimen was
exposed for a short period to a "standard atmosphere" (r.h. 95% obtained by means
of 2-5M H2SO4, with SOs generated from 0.1 ml of 0-06M K2S~Os added to 1.0 ml of
the acid); in the "standard procedure" this first stage lasted 4 h. During the second
stage, the specimen was, in general, exposed to moist air at r.h. 95% without sulphur
dioxide for a chosen time.
In both one-stage and two-stage experiments, specimens were exposed for different
periods, so as to provide corrosion-time curves (Figs. 3 and 4). In general, they were
performed in triplicate (some only in duplicate). In Sets A - D the results of individual
experiments have been reported elsewhere; z° as the results are needed for the inter-
pretation, they are reproduced in Fig. 3 in somewhat modified form.
There were eight sets of tube experiments:
Set A, One-stage experiments in moist air without SO2 (curves A of Fig. 3).
Set B, One-stage experiments under standard conditions (curves B of Fig. 3).
Set C, Two stages, with first stage (4 h) under standard conditions and second stage
in moist air without SO2 (curves C of Fig. 3).
Set D, Three stages, with first stage (4 h) under standard conditions, followed by
20 h immersion of the specimen in water, and then the final stage in moist
air without SO,, (curves D of Fig. 3)
 Mechanism of atmospheric rusting 233

90-0

67",5 4hr Ist stage

B • • C "
45'0
• "//2. : -- .
22'5
E '-- C ..... D ......
0

o~ -22"5
c
o
o - 4.5"0

-67"5

-90'0

-112"5

--~.--' ,~ ~ ~ t f t ! I
12 8 I 12 4 8 16

Time, days. Time, days

F[~. 3. Experhnents showing influence of SO.~. Curves A relate to iron exposed to

moist air without SO.., and curves B to iron in moist air with SO-"; both gain and loss of
weight are much greater when SO*- is present. Curves C represent two-stage experiments;
after 4 h in moist air containing SO2, the specimens were moved to moist air without
SO2; yet the weight--change is similar to B, showing that when once FeSO~ has been
formed, SO.. in the gas-phase is not necessary. Curve D represents three-stage experi-
ments; after 4 h in moist air with SO-", the specimens are placed in water for l d arld
then in moist air without SO..; the weight-change is small, because the FeSO~ has been
removed.

90.0
4hr Ist stage E :
67"5

45'0i

22"5 z I

oi •~ 2 2 2 : 2 : 2 - : : 2".. 2: : :.2 : 2= - : .==7." : :': : : 22.: 2

E

~t~ -22"5
g
j%
'~ - 450

01
-67"5

-90"0

-112"5

-1~5"C t I I I ~ ,
; 2 4 8 16
Time, days
FiG. 4. Experiments with replenishment of SOz. Curves E represent specimens exposed
for 4 h to moist air with SO-", and then moved to a tube containing fresh moist air with
SO-"; thus they receive a double ration of SO2. Curve B, in which the specimen receives a
single ration, is reproduced for comparison. The final effect of the double ration is nearly
to double the weight--change, although the initial effect of the extra SO= is a slight
retardation.
234 U. R. EVANSand C. A. J. TAYLOR

Set E, Two stages, with first stage under standard conditions, followed by a second
stage in tubes containing a fresh supply of KsS205, equal to that used at the
start of the first stage (curves E of Fig. 4, the broken curves repeat Set B).
Set F, Two stages, with first stage lasting 4 h, but with variation of the amount of
KsS2Os; second stage in moist air without SOs for 44 h--making a total of
48 h (curves F of Fig. 5).
Set G, Two stages, with first stage lasting for different periods with the standard
amount of KsSsOs, and the second stage in moist air without SOs to make a
total of 48 h (curves G of Fig. 6).
Set H, Two stages, with first stage under standard procedure, but second stage
conducted over HsSO4 of various concentrations (to provide various r.h.
values) without SOs (curves H of Fig. 7).

4C

30 + F

20 ÷

÷ ÷

io /
÷
/÷
0 . . . . . . . . . . . . . . . . . .

-I0
t
o
z=
-20 ~,, *

-30 +
+
÷ ~
g ~ ÷
-4o ÷

I I I I
0'01 0"02 0"06 0"18
K2S2Osin Ist stage, M

FXG. 5. Experiments with variation of the amount of $02. Curves F represent specimens
exposed to moist air containing different quantities of SOs for 4 h, and then to moist
air without SOs for 44 h. The final weight-change increases with increasing SO.~up to a
point, and then decreases, confirming Fig. 4.

The ratio of weight-loss to weight-gain obtained in Set C is calculated in Table I.

Whilst the tube experiments provide factual information about the effect of varying
different factors, they do not decide the mechanism. To arrive at such a decision, the
plate experiments were carried out.
Plate experiments. The apparatus used is shown in Fig. 8. In all cases a weighed steel
specimen measuring 6.5 × 4.0 cm was placed horizontally in a Petri dish and covered
 Mechanism of atmospheric rusting 235

4°I
30
v G

IG

0 . . . . . . . . . . . . . . . . .

7:
-2c 2

-~c !

-4C

! ' I
05 I 2

Duration of Ist stage, hr

FiG. 6. Experhnents with variation o f time in moist ah" containing SOy Curves G
represent specimens exposed for different times to moist air containing a standard
quantity of SOp., and then to moist air without SOs for a period making a total of 48 h.
The weight-change becomes constant when the first stage exceeds about 2 h--pre-
sumably the time needed to use up the SO~ in forming FeSO4.

TABLE 1. R A T I O OF WEIGHT-CHANGES IN SET C

Duration of
second stage Weight-gain Weight-loss Ratio
d (rag) (mg)
11"3 14-5 1"28
1 12"0 16"0 1"33 t 1'33
12"7 17"5 1"38

27-2 39-2 1"44

2 29'5 43"3 1"47 t I '46
29'8 43"8 1"47

40'2 61-0 1'52

4 32'5 50"0 I "54 t 1"53
38"8 59' 1 1'52

39"7 59'4 I "49

8 48'4 73"8 1"52 1'53
41-8 65-8 1"57

55'8 95'7 1"71

16 43"6 75"4 1"73 t 1'68
49"0 78"8 1"61
236 U.R.. EVANSand C. A. J. TAYLOR

40
2
30 z/"
2C /
o~
E

==
IC /
(;
z~~-3 •

.L~ -I¢
G)

-20

-3Q

-40

, , , . . , . . . ,
I0 20 30 4 0 0 7600 9 0 80
5 I00

Relative h u m i d i t y , 2nd stage

FIG. 7. Experhnents with variation of relative humidity. Curves H represent specimens

exposed to moist air with SO2 for 4 h, and then to air without SOs but at different
humidities. The weight-change rises abruptly about r.h. 70%.

with filter-paper soaked in M/10 ferrous sulphate. In Set J it was covered with plastic
(to exclude air) but in Sets I and K it was left uncovered (exposed to air). In Set K,
the filter-paper was turned upside-down every 10 min, so that any sulphuric acid
formed by oxidative hydrolysis during the 10 min preceding an inversion could have
a chance of acting on the steel during the 10 min which followed it.
In Sets L, M and N the filter-paper was covered with rusted steel, which in Set L
was not connected to the lower (unrusted) steel; in Sets M and N there was direct
electrical connection. In Set N the upper (rusted) specimen was turned upside-down
every 10 min, so that the ferric rust, after reduction to magnetite by cathodic reaction,
was periodically re-oxidized to ferric rust.
The experiments of Sets I - N were carried out in triplicate on steel, and again on
iron. Reproducibility was poor, probably for reasons suggested in the Appendix. The
measurements are shown in Table 2; 23 out of the 24 experiments in Sets I - L produced
weight-increases (the other showed no change), whilst all experiments in Sets M and
N produced weight-losses.
In Set O, the lower (unrusted) specimen was weighed, and the two were joined
through a milliammeter, current readings being taken every half-minute; the upper
(rusted) electrode was turned upside-down every 5 min, so that the current, which
had become low, at once revived; Fig. 9 presents a typical result. The area below the
zig-zag curve was measured, giving the milliampere-hours obtained. The results are
shown in Table 3.
 M e c h a n i s m o f a t m o s p h e r i c rusting 237

Plastic or rusted metal

Filter paper soaked

',,~ ~ in M/IO Fe SO4

Weighed metal anode " ~ , , ~ ~ ~

FiG. 8. Plate experhnents. A weighed metal specimen is covered with filter-paper

soaked in M/10 FeSO~ and in most experiments this is covered with plastic or rusted
metal; the latter in some cases is joined electrically to the lower specimen, either directly
or through a milliammeter. The specimen loses or gains weight in different cases--as
explained in the text. The results favour the Electrochemical Mechanism and not the
Acid Regeneration Cycle.

TABLE 2. WEIGHT-CHANGES IN m g 1N PLATE EXPERIMENTS

LASTING 1 h

Iron electrodeposited
Steel on steel basis
Set I ,-I- 1.46, + 0.83, ÷ 0.50 ÷ 0.3, + 1-0, .1. 0.7
J .1. 0.08, .1. 0.31, .1. 0.20 .1. 0.1, nil, -I- 0-3
K + 2.29, + 1.00, + 1'70 + I'1, + 1"9, + 1'5
L -t- 0.24, + 0.23, + 0-11 .1. 0.2, .1. 0.2, + 0.2
M -- 0.48, -- 1"59, -- 0.84 -- 0.8, -- 0.4, -- 0.3
N -- 4.00, -- 1.20, -- 1-18 -- 1"0, -- 0'4, -- 1.0

Weight loss of anode 2.18 mg

~ 6
E 4

I ~ i
0 5 I0 15 20 25 30 35 40 45 50 55 60

Time, min
FIG. 9. Currentpassing between rusted iron above and weighed specimen below. I n this
experiment, the upper specimen is inverted every 5 min, so that the magnetite formed by
cathodic reduction o f the ferric rust is re-oxidized by air to the ferric state. T h e milli-
a m p e r e - h o u r s are roughly equivalent to the weight-loss of the lower specimen in the
sense o f F a r a d a y ' s L a w - - a s s h o w n in Table 3.

Other experiments. Three questions vital to the interpretation of the results

required answering. They concerned:
(1) The behaviour of iron in ferrous sulphate solution, either freshly prepared or
238 U . R . EVANS and C. A. J. TAYLOR

TABLE 3. COMPARISONBETWEEN ELECTROCHEMICALMEASUREMENTSAND WEIGHT-LOSSES

Experiment number (1) (2) (3) (4)

Material Iron Iron Steel Steel
Milliampere-hours 2"35 1'95 5.0 5' 1
Corresponding to a theoretical weight-loss* (mg) 2.44 2.02 5.2 5"3
Actual weight-loss (mg) 2.18 1.97 7.0 7.0

*Assuming that only Fe 2+ ions are formed.

pre-exposed to air, (2) the nature of the products of oxidative hydrolysis, and (3) the
behaviour of rusted iron in various solutions--including ferrous sulphate.
Two sets of experiments sufficed, since Set P answered two questions:
Set P. Shallow pools of freshly prepared M/10 ferrous sulphate solution (about
28 ml each) were introduced into Petri dishes (about 8.5 cm dia.), and exposed to air
over water at 25°C for various periods (8, 4, 2 and 1 d); it was arranged for all these
periods to end on the same day, and on that day a fresh solution was also prepared.
The solutions were filtered, and allowed to act, in tubes, on weighed iron specimens.
The change produced was recorded, as represented by
(1) evolution of hydrogen gas, (2) change in appearance after 70 min and after
5 h, and (3) change in weight after 70 min and 5 h.
It was found that the liquid exposed to air produced bubbles of hydrogen after
a few minutes; the specimens were visibly corroded when taken out, and were found
to have lost weight. The attack, as judged by all three criteria, tended to decrease
as the period of preliminary air exposure became shorter; the liquid exposed for 1 d
produced very little change of weight or appearance. Freshly prepared solution
produced practically no visible change, and there was a slight gain in weight.
The Petri dishes carried an adherent rust-like deposit, evidently the result of
oxidative hydrolysis. This was washed until the wash-water showed no trace of Fe ~÷
or Fe 3+ when tested; it was then dissolved in dilute hydrochloric acid, and the solution
was found to contain SO42-. Evidently the product of oxidative hydrolysis is not entirely
FeOOH, but contains basic sulphate--a matter of some importance, as shown later.
Set Q. Steel carrying adherent ferric rust was placed in tubes completely filled with
various liquids without air-space; liquids tested included M/10 ferrous sulphate,
M/10 sodium sulphate, M/10 sodium chloride, M/10 ammonium chloride and distilled
water. Ferrous sulphate produced darkening of the rust after 2 min; the surface was
fully black after 20 rain, by which time no darkening had been produced by either
sodium salt or by distilled water; slow darkening did occur in these liquids after some
hours, but that is probably attributable to the slow formation of small amounts of
ferrous compounds by attack on the iron; this explanation is supported by the
observation that ammonium chloride, which is known to attack iron more quickly
than sodium chloride, produced appreciable darkening of the rust after 30 min.

INTERPRETATION
Relation between weight-gain and weight-loss. The pairs of curves in Figs. 3-7 are
not mirror-images. If the corrosion-product consisted only of FeOOH, the weight-loss
should always be 1.69 times the weight-gain (the value of Fe/OOH). Table 1 shows
 Mechanism of atmospheric rusting 239

that for short experiments the ratio is lower; it practically reaches the theoretical value
after 16 d. The early disagreement is easily understood, since the product at first con-
tains much ferrous sulphate. The final agreement may seem impressive, but in fact
traces of compounds other than FeOOH are still present, even after 16 d; probably
some of these tend to raise and others to lower theratio, so that in effect it comes close
to the theoretical value.
Importance of ferrous sulphate in causing continued rusting. Figure 3 confirms the
earlier observations n that, when once ferrous sulphate has appeared, the presence of
sulphur dioxide in the gas-phase is not needed for rusting. The specimens (curves C)
moved after 4 h to moist air free from sulphur dioxide give practically the same
performance as those left in the original atmosphere (curves B); but the specimens
from which ferrous sulphate has been removed by a day in water (curves D) suffer
practically no rusting, behaving more like specimens exposed from the outset to moist
air free from sulphur dioxide (curves A). This interpretation is confirmed by the
curves G of Fig. 6, where the duration of the preliminary exposure to sulphur dioxide
is varied; if this exceeds a certain value (evidently representing the time needed to
convert all the sulphur dioxide to ferrous sulphate), further lengthening has no effect
on the rust produced after a total period of 48 h. If, after the standard time in moist
air with the standard amount of sulphur dioxide, a second "ration" of sulphur dioxide
is provided (curves E of Fig. 4), the final effect is almost to double the total amount of
rust produced; this is easily understood, since the amount of ferrous sulphate formed
will be doubled. However, the immediate effect of the second addition of gulphur
dioxide is slightly to diminish the rusting-rate, and this change, although small, is
probably significant, since curves F (Fig, 5), where the amount of sulphur dioxide
present in the first stage of two-stage experiments is varied, show that the rusting
after a total time of 48 h first increases with the amount of sulphur dioxide present,
but that large amounts slightly depress the amount of rusting. The reason is probably
that, when the amount of sulphur dioxide is excessive, a second anodic reaction, the
oxidation of SO]- to SO]- can compete with the main anodic reaction (Fe --~ Fe2+),
so that the amount of iron attacked now becomes less than the equivalent of the
cathodic reduction of oxygen.
All the curves of Figs. 3 and 4 seem to become nearly asymptotic to a line parallel
to the time-axis; this suggests that a given amount of ferrous sulphate can produce only
a finite amount of rust, although they confirm Schikorr's observation 8 that one
molecule of SO2 ~vill produce many molecules of rust. This is not inconsistent with the
Acid Regeneration Cycle. The Electrochemical Cycle, if it operated alone, would
predict that a single molecule of FeSO4 could lead to an infinite number of molecules
of FeOOH; it is shown later that hydrolysis, by removing the Fe 2÷ and SOl- as a
basic sulphate, causes the amount of rust formed to be finite.
Distinction between the two mechanisms. The plate experiments of Sets I-O were
designed to decide whether Acid Regeneration or the Electrochemical Cycle is most
important.
Set K represented conditions most favourable to the Acid Regeneration mechanism.
The ferrous sulphate on the upper side of the filter-paper exposed to air should be
oxidized and hydrolyzed to form sulphuric acid, which, when the filter-paper is
inverted, comes into contact with the iron, and may give fresh ferrous sulphate; the
240 U.I~EvANS and C . A . J . TAYLOR

latter will only be converted to FeOOH 10 min later, when the paper is once more
inverted; the FeOOH, formed out of contact with the metal, should not have any
protective action. No adherent product would be expected on the iron, and there
should be a loss of weight. When there is no access of oxygen to the metal, as in Sets
J and L, these events should not take place and no weight-change would be expected. In
Set I, oxygen diffusing through the filter-paper might cause the changes mentioned to
proceed close to the metallic surface; thus some adherent product might be expected
and it cannot be predicted whether the weight ought to increase or decrease. The
experimental results, however, show a weight-increase in all four sets (Table 2), and
this is greatest in the case of Set K, where, if Acid Regeneration were operative,
weight-loss would definitely be predicted. The results are, therefore, unfavourable
to the Acid Regeneration theory.
The weight-gain is probably due to evolution of hydrogen--which is likely to
occur if oxygen is excluded, as in the plate experiments. Reactions leading to hydrogen-
production are not without interest, but they probably play little part in atmospheric
corrosion under ordinary conditions; the discussion of such reactions has, therefore,
been relegated to an Appendix.
Sets M and N were designed to provide conditions favourable to the Electro-
chemical Cycle. If this is operative, weight-loss would be expected, and is in fact
observed. Set O (Fig. 9) show that the current produced by the cell Fe[FeSO~ [ FeOOH
soon weakens with time, evidently because the FeOOH is used up; but when the upper
specimen is turned upside-down, so that a fresh supply of FeOOH (produced by air-
oxidation of the Fe~O4 formed during the previous cycle) becomes available, a strong
current is once more obtained. Figure 9 shows the cycle (cathodic reduction of FeOOH
to Fe304, followed by its reconversion to FeOOH by air-exposure) repeating itself
eleven times; in other experiments it was found possible to obtain fresh current 15
times without any sign that the regeneration was failing.
Each milliampere-hour should, by Faraday's law, correspond to the corrosion of
1-04 mg of iron, if the only anodic reaction is

Fe ---- Fe 2+ q- 2e,

but to two-thirds of that amount if every Fe ~+ ions is anodically oxidized to Fe a+

before it escapes from the metal surface

Fe 2+ = Fe 8+ q- e.

In the corrosion of iron vertically immersed in a salt solution, there will usually be
complete escape, and under such conditions, Hoar 1 found good agreement with the
weight-loss calculated on the assumption that only Fe 2+ is formed (in his experiments
chemical oxidation to the ferric condition occurred out of contact with the metal).
For a horizontal surface covered with filter-paper, or, as in real atmospheric rusting,
with a rust layer, escape will be less easy, and some at least of the Fe ~+ may be expected
to be converted electrochemically to Fe s+. Thus Thornhill, ~1 operating on surfaces
covered with soaked filter-paper, obtained figures intermediate with those calculated
on the two assumptions (of Fe ~+ and Fe 3÷ as sole products). The same is true of the
 Mechanism of atmospheric rusting 241

results recorded in Table 3. In the two experiments with iron the measured weight-
losses are definitely less than those calculated on the assumption that the sole anodic
product is Fe 2+; presumably there is some electrochemical production of Fe 3+. In the
case of steel, the weight-loss as measured is rather more than that calculated from the
electrochemical observations. This is probably due to the fact that, where a second
phase, such as cementite, is present, the anodic attack will fall preferentially on the
iron atoms at the boundary with that second phase, where atomic disarray is to be
expected; consequently, some of the second-phase particles may be dislodged un-
corroded in a manner unconnected with the evolution of electric current.
No exact correspondence between electrochemical and gravimetric measurements is
obtained. In view of the scatter shown in Table 2, and the various complicating factors
discussed in the Appendix, exact agreement cannot be expected. But the current
measurements obtained are of the right order of magnitude, and no mechanism--
other than the electrochemical m e c h a n i s m - c a n explain the results. It is evident,
however, that the picture presented in the introduction to this paper represents an
over-simplification. In addition to the two complicating factors just mentioned (the
formation of Fe 3+ ions and the dislodgement of particles uncorroded), it should be
recollected that the compositions of the three oxides of iron, including magnetite,
can vary over a range of values--as shown by Pfeil ;22 magnetite does not always have
the composition exactly represented by the formula FeaO4.
The fact that Set K, where the filter-paper was periodically turned upside-down,
gives a greater increase of weight than Sets J and L, where air is excluded, is easily
explained. After each inversion dissolved oxygen will reach the metal, and produce
adherent product, thus adding to the weight-gain. This would be expected, even
if the reaction consisted in direct chemical combination of iron and oxygen--which
is unlikely. A weight-gain would equally be expected, if the formation of rust depends
on the Electrochemical Cycle, since here the cathodic and anodic reaction will occur
on the same specimen but at different points. In Sets N and O the anodic and cathodic
reactions occur on different specimens; weight-losses of the anodic specimens are to
be expected and are, in fact, observed.
The results as a whole support the Electrochemical Cycle and not the Acid
Regeneration Cycle. To ascertain why the latter fails, Set P was carried out, in which
shallow layers of ferrous sulphate solution were exposed to air in Petri dishes at 25°C
for various periods and then filtered; the filtrate was allowed to act on weighed iron in
a tube. The solution which had been exposed to air for 8 d attacked iron with evolution
of hydrogen-gas, producing a loss of weight. In contrast, freshly-prepared ferrous
sulphate solution showed practically no visible change and a slight gain in weight. If
the air-exposure had lasted one day only, there was again practically no visible change,
and the loss of weight was smaller thanwith 8-d exposure. It is evident that oxidative
hydrolysis, liberating acid and rendering the liquid corrosive--as postulated by the
supporters of the Acid Regenerative Cycle--does in fact take place, but very slowly.
The Acid Regeneration Cycle requires days for its completion, whereas the Electro-
chemical Cycle can be repeated several times within 1 h. It is not surprising, therefore
that the former contributes only a negligible positive effect to atmospheric corrosion.
It has, however, an important negative effect. The rusty layer left on the Petri
dishes after the air-exposure of the ferrous sulphate solution was washed until the
242 U . R . EVANS and C. A. J. TAYLOR

wash-water ceased to show the reactions of Fe z+ and Fe3+; it was then dissolved in
hydrochloric acid and tested with barium chloride. Much SO42+ was found, and it is
evident that the rust formed by oxidative hydrolysis is not merely FeOOH but also
contains a basic sulphate. Thus there is removal of Fe z+ and SO~- from solution,
which will reduce the conductivity of the liquid joining anodic and cathodic points
and slow down the electrochemical reaction. This explains why a finite quantity of
ferrous sulphate does not produce an infinite amount of rust--as would be expected
if the equations suggested represented the whole story.
It appears, therefore, that the Acid Regeneration process--which has been regarded
as the most important factor in producing corrosion--is really important in limiting
the amount of corrosion which the Electrochemical Cycle would be expected to
produce.
The results of Set Q support the idea that the conversion of hydrated ferric oxide
to magnetite results from a cathodic reaction involving the entry of Fe z+ into FeOOH.
They seem to rule out an alternative mechanism, not requiring Fe z+ in the liquid, but
assuming the discharge of H ÷ ions, which would leave O H - in excess, and precipitate
Fe(OH)z, giving magnetite by interaction between Fe(OH)z and FeOOH; if this occurs
at all, it takes place much more slowly than the cathodic reaction involving Fe z+.
Outdoor and indoor conditions. The manner in which the two stages irt the Electro-
chemical Cycle contribute to attack will depend on conditions. If the iron is exposed

Air
Fe 0 OH with ~v t / , , , ' J - s - t - ¢ , ~ , t / ~ , t . ' , t , ~ . ~ t
air in channels , - c ~ .'~,-C.,~.';,,",'...~t I , , , - t ' ~ ; ~ t ~ ~,-

Fe304 with LP# &AP ~ LD # ~ ~OLp 8 Ib L" AD ~ L ~ # •

Fe SO 4 solution ~q= Q J ~ - - D -D= -~=--* eo~'~
in channels X'
X Metal

FIG. 10. Probable situation o f iron developing rust in moist air but protected from rain.
Anodic attack on the iron at X X ' is balanced by cathodic reduction of the rust at Y Y '
to magnetite which, however, is quickly re-oxidized by air to the ferric state. (Note: The
situation will vary with conditions. A rust-coat on specimens exposed out-of-doors will
become largely converted to black magnetite during rain-spells, but this will be largely
re-oxidized to brown ferric rust during intervening dry weather.)

out of doors in a position where it remains covered with liquid during a rain-storm,
and then dries up, the brown rust will probably be reduced to black magnetite under
rain (as shown experimentally in Set Q) and then during the suceeding dry period
the black will be re-oxidized to brown rust. Thus the two stages of the process will
occur alternately--as in the experiments of set N. If, however, the iron is exposed
indoors, a steady state will be established, probably in the manner suggested in Fig. 10;
anodic attack on the iron will occur at a level XX', producing Fe z+, but the same
amount of Fe z+ will be removed at level YY', entering the rust by the cathodic reaction
and reducing it to FesO4, which will at once be re-oxidized by the air to FeOOH.
Since Fe z+ is formed at one place and removed from solution at another, the solid
 Mechanism of atmospheric rusting 243

produced will not be protective, explaining why rusting continues (on iron free
from alloying constituents) at a nearly constant speed. 4
To maintain corrosion the cell requires (1) an electronically conducting limb, the
magnetite, and (2) an electrolytically conducting limb, the ferrous sulphate solution.
Corrosion will become slow if the resistance of either limb is high. In the case of iron,
electronic conductivity is provided by magnetite, but most other metals form no
intermediate oxide possessing good electronic conductivity. This is doubtless the
reason why zinc, under normal atmospheric conditions, suffers change far more
slowly than iron, so that a layer of zinc is commonly used to protect iron--despite the
fact that the free--energy drop involved in conversion to oxide or hydroxide is greater
for zinc than for iron. Only in highly acid atmospheres is zinc attacked more seriously
than iron; Hudson 6 found that in a railway tunnel the attack on zinc was faster than
that on iron. Under immersed conditions, the same may be true even of neutral
solutions; early measurements ~3on plates partly immersed in chloride solutions showed
greater corrosive velocity for zinc than for iron. This is easily understood.
In an industrial atmosphere, the electrolytic limb is provided by the ferrous sulphate
solution. If the r.h. declines below the critical value, so that the solution largely dries
up, current will almost cease to flow and corrosion will become very slow--as is found
to be the case.
In this research no work has been carried out on alloy steels. The causes of the
protective action of the invisible film on stainless steel were examined at Cambridge
some years ago. z~,25 The protective action of the visible rust-layer formed slowly on
low-alloy steels has recently received careful studies in Japan z~,27 and elsewhere, zs

APPENDIX
Cause of unexpected weight-gains and poor reproducibility. In certain experiments
(Sets I-L, P) ferrous sulphate solution produced unexpected weight-gains; these
generally appeared under conditions where oxygen was excluded or restricted, and
probably represent a phenomenon not likely to occur to any great extent in atmospheric
corrosion. However, they cofiaplicate interpretation of the results, and, if unexplained,
would arouse doubts. The specimens showing weight-gains generally, but not in-
variably, exhibit a visible stain, which is sometimes greenish-white when the specimen
is taken out from the vessel, but becomes brown on exposure to air. The phenomenon
is associated with a scatter of results much greater than that met within the experi-
ments of Sets A - H ; an explanation of this poor reproducibility must be provided if
the results are to avoid suspicion. In fact, however, it is to be expected on the basis of
present knowledge that, in absence of oxygen, ferrous sulphate would act on iron,
giving an adherent product to an extent which will vary from one specimen to another.
It has long been known that reaction of iron and water with liberation of hydrogen
is a thermodynamically possible reaction. The situation is conveniently expressed in
diagrams due to Pourbaix, 29 which shows that the driving force, expressed as the
e.m.f, of the cell Fe]H2 is 47 mV if the solid product is Fe(OH)8 and 85 mV if it is
FezO4. Experiments have confirmed theoretical expectation. Murata s0 found that water
can act on iron powder, liberating hydrogen and producing ferrous hydroxide.
De Kay Thompson sl observed a similar reaction, but with magnetite as end-product.
There is no necessary inconsistency here, since ferrous hydroxide, as shown by
244 U.R. EVANSand C. A. J. TAYLOR

Schikorr, s2 can itself react, liberating hydrogen and producing a ferrosoferric com-
p o u n d - p r e s u m a b l y magnetite. Now Wanklyn a3 has shown that the Schikorr reaction
does not occur with pure materials, although it proceeds readily when copper powder,
nickel powder or colloidal platinum is present. Presumably we are concerned with an
electrochemical change on the metallic surface, with hydrogen liberated by the
cathodic reaction at favourable points, whilst at other places, the anodic reaction
Fe 2+ -----Fe ~+ q- e is occurring. So soon as a small amount of Fe 3+ has been formed,
magnetite, which is much less soluble than ferrous hydroxide, will separate. Evidently,
in Murata's work, impurities favourable to the Schikorr reaction were excluded from
the reaction vessel (but the vessel used for measuring conductivity contained platinum
electrodes, and some hydrated magnetite was formed on the platinum).
In absence of electrolyte the action of salt-free water is found to be slow, as is
easily understood. A simple chemical reaction (or an electrochemical reaction with
anodic and cathodic changes occurring over the whole surface) must lead to a solid
film in physical continuity with the surface, which, at ambient temp., is likely to slow
down the attack. If there were electrochemical action with the anodic and cathodic
reactions occurring at discrete points, the stifling of the attack by a film could be
avoided, but it would leave unbalanced Fe 2+ and O H - at the anodic and cathodic
points respectively; such a change could not long continue.
If, however, the liquid contains ferrous sulphate, the reaction should proceed
without hindrance. Fe 2+ will be formed at the anodic points, whilst at the cathodic
points, the OH + left when hydrogen is liberated will combine with Fe z+ to give solid
ferrous hydroxide; thus a film will be formed, but it will not, at least for some time,
cover the anodic points, and the reaction does not involve the formation of unbalanced
charged particles; as much Fe z+ is removed at the cathodic points as is produced at the
anodic points, and the adherent matter must lead to an increase of weight.
The Pourbaix diagram 29 shows that over a wide range of pH e.m.f, providing
the driving-force (for the production of either Fe(OH)2 or Fe304) is constant; but
in acid solution the e.m.f, available is much greater and the product is then a soluble
substance, so that more rapid attack leading to a weight-loss must be expected.
This accords exactly with what is observed. A freshly-prepared solution of ferrous
sulphate acts slowly on iron, causing a weight-gain. A solution which has been exposed
to the air sufficiently long for oxidative hydrolysis, acts more rapidly, giving for a time
a marked evolution of gas, and causing a weight-loss.
The fact that hydrogen is evolved explains the poor reproducibility. It is common
knowledge that hydrogen-evolution generally occurs from a limited number of points,
and frequently a few points on a specimen will be noticed to provide much more
hydrogen than others. On a small specimen, there may be, on an average, only a few
oftheseimportant points, and probability principles then predict poor reproducibility.
If, for instance, the average number on a specimen is two, there is a fair chance that
certain specimens may have four, three, one or even no points of that character, and
the latter specimens will react much more slowly than the former. The fact that poor
reproducibility is obtained in corrosion reactions in which hydrogen is liberated, has
long been recognized. An early example may be quoted, although probably a study
of the literature would reveal still earlier ones. Experiments on the attack of zinc to
hydrochloric acid, published in 1925, showed that identical specimens, cut from care-
 Mechanism of atmospheric rusting 245

fully produced zinc sheet, yielded different curves, although the general shape was
the same. a4
The lack of agreement between the duplicate experiments in Sets I - O need not,
therefore, be regarded as a sign that these experiments were less satigfactory t h a n
Sets A - H . It does, of course, mean that little or no significance can be attached to tile
numbers recorded, but the fact that with one exception all the weight changes in
Sets I - L are gains and all those in Sets M and N are losses c a n n o t be ascribed to
chance; this conclusion has been confirmed by the signs of the changes obtained in
other experiments carried out under slightly different conditions.
In Set O, where comparison between weight-loss a n d milliampere-hours is
important, the value for the total milliampere-hours represents the average of twelve
5-rain periods, so that, despite the serious differences in the curves for the different
periods (see Fig. 9), the average should represent a significant f i g u r e - - i n accordance
with the accepted principles of error-theory. Thus the c o m p a r i s o n between calculated
and observed weight-losses (Table 3) has significance, although agreement is n o t
exact. It is not clear why, in an experiment conducted on the same pair of electrodes,
the shape of the curves for different periods should vary so much. A n investigation
of the cause would possess interest, but would probably have no bearing on the
mechanism of Atmospheric Corrosion.

Acknowledgemet,ts--Thanks are due to the Ministry of Defence (Aviation Supply) for support of the
work, and particularly to Mr. H. G. Cole and Mr. C. E. Hoey for interest and helpful advice; also
to Professor R.. W. K. Honeycombe and Dr. J. E. O. Mayne for permission to carry out the work in
the Department ; finally to J. F. Stanners, L. L. Shreir and G. Butler for useful comments.

REFERENCES
1. U. K. EVANSand T. P. HOAR,Proc. R. Soc. 137(A), 343 (1932).
2. F. BLAHA,N~.tnre, Lond. 166, 607 (1950).
3. U. R. EVANS, Trans. Faraday Soc. 19, 201 (1923).
4. W. H. J. VERNON,Trans. Faraday Soc. 19, 886 (1924); 23, 159 (1927); 31, 1668, 1678 (1935).
5. A. BUr.OWIECKI,Schweizer Archiv Wiss. Tech. 23, 97 (1957); 32, 42 (1966).
6. J. C. HUDSON,reports to the Iron and Steel Institute Corrosion Conu-nittee from 1930 onwards;
J. C. HtrDSONand J. F. STANNERS,J. appl. Chem. 3, 86 (1953).
7. K. A. CHANDLERand M. B. KaLCULLEN,Br. Corros Jnl. 3, 80 (1968).
8. G. SCHIKORR, Werk~toffe Kcrros. 14, 69 (lC63); 15, 457 (1964); 18, 514 (1967).
9. W. H. J. VERNON,Chem. Ind. 62, 325 (1943); J. Roy~ Soc. Arts 97, 589 (1949).
10. U. R. EVANS,Trans. Inst. Met. Finishing 37, 1 (1960); esp. p. 5.
11. U. R. EVANS,Natl, re, Lond. 206, 980 (1965).
12. H. SCHWARZ, Werkstoffe Korros. 16, 93,208 (1965).
13. G. SCHIKO~, Private Communication about 1965.
14. K. BARTONand Z. BArt'rONOV.~, Werkstoffe Korros. 20, 216 (1969).
15. R. S. THORN~XLL,Private Communication (1952).
16. A. G. TANNER,Chem. hid., p. 1027 (1964).
17. T. K. Ross and B. G. CALLAC;HAN,Corros. Sci. 6, 377 (1966).
18. D. FYFE, C. E. SnANAHANand L. L. SHREm, Corros. Sci. 8, 349 (1968).
19. M. J. PRYORand U. R. EVANS,J. chem. Soc., p. 3330 (1949); pp. 1259, 1266, 1274 (1950).
20. U. R.. EVANS, Corros. Sci. 9, 813 (1969), esp. Fig. 2, pp. 818-9.
21. R. S. THOR~qlLLand U. R. EVANS,J. chem. Soc., p. 614 (1938).
22. L. B. FFEIL,J. Iron Steel Inst. 119, 501 (1929); 123, 237 (1931).
23. U. R. EVANS,J. chem. Soc., p. 111 (1929), esp. Fig. 1, p. 124.
24. I. D. G. BERWlCl¢and U. R. EVANS,J. appL Chem. 2, 576 (1952).
25. U. R. EVANS, Chem. Ind., p. 1779 (1962).
26. I. MATStlSmMAand T. UENO, Corros. Sci. 11, 129 (1971).
27. T. MISAWA,T. KYUNO,W. SUETAKAand S. SmMODAmA,Corros. Sci. 11, 35 (1971).
246 U..R.. EVANS and C. A. J. TAYLOR

28. D. FY~, C. E. A. SHANAr~ANand L. L. SHREIR,Corros. Sci. 10, 817 (1970).

29. M. POURBAIX,Atlas of Electrochemical Equilibria in Aqueous Solutions, pp. 290, 312. Pergamon,
Oxford (1966).
30. K. MURArA, J. Soe. chem. Ind. Japan 35, 525B (1932).
31. M. DE KAY THOMPSON, Trans. electrochem. Soc. 78, 251 (1940).
32. G. SCmKORR, Z. Electrochem. 35, 65 .(1929); Z. anorg, allg. Chem. 212, 33 (1933).
33. J. N. WANKL~ and U. R. EVANS, Nature, Lond. 162, 27 (1948).
34. U. R. EVANS,J. lnst. Metals 33, 27 (1925); esp. Fig. 1, p. 39.