You are on page 1of 26

Corrosion Science 52 (2010) 455–480

Contents lists available at ScienceDirect

Corrosion Science
journal homepage: www.elsevier.com/locate/corsci

Potential and current distributions of one-dimensional galvanic corrosion systems


Guang-Ling Song *
Chemical Sciences and Materials Systems Laboratory, Research and Development Center, GM, Mail Code: 480-106-212, 30500 Mound Road, Warren, MI 48090, USA

a r t i c l e i n f o a b s t r a c t

Article history: Many practical galvanic corrosion problems can be simplified into a one-dimensional mathematical
Received 21 June 2009 equation. In this study, theoretical expressions for galvanic potentials and currents of one-dimensional
Accepted 7 October 2009 systems are deduced and some critical parameters involved in the potential and current equations are
Available online 6 November 2009
systematically discussed. The developed analytical approach is then applied to some practical galvanic
corrosion cases, such as a steel–Al joint exposed to bio-fuel E85, a galvanic couple separated by a passive
Keywords: spacer and a scratched organic coating. It is demonstrated that the analytical approach can examine the
Galvanic corrosion
reliability of a computer modeled galvanic corrosion process and help understand the mechanism behind
Modeling
Linear system
the computer modeled galvanic corrosion behavior.
One dimension Ó 2009 Elsevier Ltd. All rights reserved.

1. Introduction Unfortunately, due to the complexity in geometric shape of


some mixed-metal components, predicting the galvanic current
Galvanic corrosion damage is common to components with and its distribution over a complicated component is actually dif-
mixed metals [1–5]. In some cases a multi-phase metallic mate- ficult. Sometimes, it is impossible to have an analytical solution
rial suffers from a severe corrosion attack simply due to the gal- to the governing equation for a galvanic corrosion system that
vanic effect between its different phases [6–8]. Moreover, in a has a complicated geometric shape and/or contains more than
sacrificial cathodic protection system, the protection current in one galvanic couple. In fact, apart from some very simple linear
effect is also a galvanic current [9,10]. In theory, the degree systems, no theoretical potential or current expression has been
and the distribution of galvanic corrosion damage can be ex- deduced so far for a complicated galvanic system.
pressed as an overall galvanic current and a distribution of the The galvanic current for a complicated galvanic system may be
galvanic current density. Therefore, reliably estimating the gal- predicted by computer modeling. However, the modeling is in nat-
vanic current density or its distribution of a practical galvanic ure a ‘‘computer experimental process”. It can only demonstrate the
system is of great significance. influence of the variation of input parameters on the output results.
Some efforts have been made in this area since the 1950s However, to fully understand a system, clearly defined analytical
[11–22]. It has been generally accepted that [23]: (1) Laplace equa- relationships between experiential parameters and experimental
tion can apply as a governing relationship for the potential distri- phenomena are essential. Thus, computer modeling cannot replace
bution in a galvanic couple system; (2) linear kinetics can be the merit of an analytical solution.
assumed to simplify the mathematic problems if corrosion rates The author believes that many practical galvanic corrosion sys-
are very low; (3) coupled metals may be treated as infinite or tems can be reasonably simplified into a one-dimensional model,
semi-infinite surfaces for simplified mathematic analyses; and of which an analytical description of the galvanic current will be
(4) there is a tendency of taking a numerical approach to estimate possible. At least two types of galvanic systems can be treated in
the galvanic current distribution for complicated practical systems. one dimension (referring to Fig. 1): (1) a fine tube containing elec-
Recently, an attempt was made to estimate the galvanic current trolyte where its cross-section area is far smaller than its length,
distribution of a Mg alloy in contact with steel by using a boundary and (2) a surface covered by a thin electrolytic film where the
element model (BEM) approach with non-linear polarization thickness of the film is much smaller than the length of the electro-
curves as boundary conditions [24] and it was found that a bound- lyte coverage. In these two systems, the radius (r) of the tube or the
ary condition closely matching a non-linear polarization curve thickness (d) of the electrolyte film is so small as to be neglected.
yielded a more accurate or reliable calculated distribution of gal- Also, the perimeter of the tube or the width of the electrolytic film
vanic current density curve than a linear boundary condition [25]. does not need to be considered. These two types of simplified one-
dimensional models represent a large number of practical galvanic
corrosion systems, such as a panel with a organic coating being af-
* Tel.: +1 586 986 1339; fax: +1 586 986 9204.
E-mail address: Guangling.song@gm.com
fected by electrolyte, a panel exposed in atmosphere with some

0010-938X/$ - see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.corsci.2009.10.003
456 G.-L. Song / Corrosion Science 52 (2010) 455–480

Nomenclature

a length of material (a), which can the length of a material IbF Faradic current density over the material (b) surface in
of the electrolyte on the material surface the direction vertical to the material surface
b length of material (b) inserted between material (a) and IbF0 Faradic current flowing in the liquid film along the
material (c) material (b) surface at the left end
c length of material (c), which can the length of a material IbFb Faradic current flowing in the liquid film along the
of the electrolyte on the material surface material (b) surface at the right end
C the inner perimeter of a studied tube system IcF Faradic current density over the material (c) surface in
m length of studied electrolyte over material the direction vertical to the material surface
m0 length of studied electrolyte over the left ‘‘dead end” of IcF0 Faradic current flowing in the liquid film along the
a multi-piece joined system material (c) surface at the left end
m0 length of studied electrolyte over the right ‘‘dead end” of E the potential relative to a reference electrode
a multi-piece joined system Ea the potential over material (a) relative to a reference
mi length of studied electrolyte over the ith ‘‘open ends” electrode
piece of a multi-piece joined system Eb the potential over material (b) relative to a reference
r inner radius of a metal tube containing electrolyte electrode
S cross-section area of electrolyte over a material surface Ec the potential over material (c) relative to a reference
or in a metal tube electrode
y width of electrolytic film over a material surface Ecorr the corrosion potential or open-circuit potential relative
d thickness of electrolytic film over a material surface to the reference electrode
i0 the current flowing into the thin liquid film from the left Eacorr corrosion potential of the material (a) relative to a
end of the section reference electrode
im the current flowing out of the thin liquid film from the Eccorr corrosion potential of the material (c) relative to a
right end of the section reference electrode
ig galvanic current flowing into or out the thin liquid film Ebcorr corrosion potential of the material (b) relative to a
at a boundary reference electrode
iag galvanic current flowing into or out the thin liquid film W the potential in the electrolytic film relative to the
over material (a) at a boundary potential of material
ibg galvanic current flowing into or out the thin liquid film Wa the potential in the electrolytic film over material (a)
over material (b) at a boundary relative to the potential of material
icg galvanic current flowing into or out the thin liquid film Wb the potential in the electrolytic film over material (b)
over material (c) at a boundary relative to the potential of material
iac
g galvanic current flowing between the thin liquid films Wc the potential in the electrolytic film over material (c)
over materials (a) and (c) at their interface relative to the potential of material
iab
g galvanic current flowing between the thin liquid films Wa0 the potential in the electrolytic film relative to the
over materials (a) and (b) at their interface potential of material (a) at boundary x = 0
ibc
g galvanic current flowing between the thin liquid films Wb0 the potential in the electrolytic film relative to the
over materials (c) and (b) at their interface potential of material (b) at boundaries x = 0
ip polarization current flowing into or out the thin liquid Wc0 the potential in the electrolytic film relative to the
film at a boundary potential of material (c) at boundary x = 0
If current density flowing in the liquid film along the W0 the potential in the electrolytic film relative to the
material surface material potential at the left end of the section
Iaf current density flowing in the liquid film along the Waa the potential in the electrolytic film relative to the
material (a) surface potential of material (a) at boundary x = a
Iaf0 current density flowing in the liquid film along the Wbb the potential in the electrolytic film relative to the
material (a) surface at the left end potential of material (b) at boundaries x = b
Iafa current density flowing in the liquid film along the Wcc the potential in the electrolytic film relative to the
material (a) surface at the right end potential of material (c) at boundary x = c
Ibf current density flowing in the liquid film along the Wm the potential in the electrolytic film relative to the
material (b) surface material potential a the right end of the section
Ibf0 current density flowing in the liquid film along the Rp polarization resistance
material (b) surface at the left end Rap polarization resistance of material (a)
Ibfb current density flowing in the liquid film along the
Rbp polarization resistance of material (b)
material (b) surface at the right end
Icf current density flowing in the liquid film along the Rcp polarization resistance of material (c)
material (c) surface Rs solution resistance
Icf0 current density flowing in the liquid film along the qs solution resistivity in the one-dimensional system
material (c) surface at the left end qp polarization resistivity in one-dimensional system
Icfc current density flowing in the liquid film along the q0s solution resistivity in a conventional three-dimensional
material (c) surface at the right end system
IF Faradic current density in the direction vertical to the q0p polarization resistivity in a conventional three-dimen-
material surface sional system
IaF Faradic current density over the material (a) surface in qap polarization resistivity of material (a) in the one-dimen-
the direction vertical to the material surface sional system
IaFa Faradic current flowing in the liquid film along the qbp the polarization resistivity of material (b) in the one-
material (a) surface at the right end dimensional system
G.-L. Song / Corrosion Science 52 (2010) 455–480 457

qcp the polarization resistivity of material (c) in the one- Lc characteristic distance of material (c)
dimensional system W distribution characteristic number
L characteristic distance Wa distribution characteristic number of material (a)
La characteristic distance of material (a) Wb distribution characteristic number of material (b)
Lb characteristic distance of material (b) Wc distribution characteristic number of material (c)

condensed liquid on its surface, a panel with water splashed over variables or parameters (e.g. corrosion potential or polarization
its surface, a water or oil pipeline, a hemmed edge of a vehicle body resistance) will not vary significantly along the width (y). For a
closure with moisture accumulated in the hem crevice, a fuel or tube, this means that, if radius (r) is very small, i.e., r ? 0, there will
brake oil pipeline in a vehicle, coolant and oil paths in an engine be no need for considering the variation of the state variables or
block, etc. parameters over the cross-section. In such a one-dimensional sys-
In this paper, we will first deduce general theoretical potential tem, the polarization resistivity qp and solution resistivity qs have
and current expressions for a uniform one-dimensional system. their definitions (qp = polarization resistance  length, and
To tackle practical galvanic corrosion problems, after the mathe- qs = solution resistance/length) and units (X cm1 and X cm) dif-
matical modeling is successfully applied to a galvanic couple with ferent from those in a three-dimensional system (see Appendix
simply an anode and a cathode joined together, it further extended A). Because of the above simplifications, the theoretical current
to a more complicated galvanic system with three different metals and potential for a one-dimensional system can be deduced as
joined together, which represents more realistic galvanic corrosion follows.
cases in practice. The theoretical galvanic current expressions are
then used to analyze the effect of a passive isolator on the galvanic 2.1. General potential and current of a one-dimensional system
corrosion of a galvanic couple, the possible corrosion damage to an
Al engine cylinder exposed to a new bio-fuel E85, and the corrosion It is assumed that the distribution of potential over a given sec-
risk along a scratch made in an organic coating. tion of a one dimensional system along x-coordinate is a curve as
shown in Fig. 2. In this simplest case, the Faradic reaction at the
interface between the electrolyte and the metal is linearly con-
2. One-dimensional primary potential and current densities trolled by the potential W in the electrolytic film (W is the electro-
lyte potential relative to the corrosion potential Ecorr of the metal).
Fig. 1 schematically shows a one-dimensional tube and flat sur- The potential distribution can be easily obtained [26] (see
face. They are not necessarily straight or flat in practice. The sim- Appendix B):
plification is made under the following conditions: x  x
E ¼ Ecorr  A exp  B exp  ð1Þ
L L
(1) The cross-section area (S) of the electrolyte must be much
where
smaller than its studied length (m). sffiffiffiffiffiffi
(2) The state variables (e.g. surface potential) over the cross-sec- qp
L¼ ð2Þ
tion along either dimension perpendicular to the length qs
direction are relatively uniform or constant compared with
those along the length direction. Correspondingly, we have expressions for Faradic current density IF
and non-Faradic flowing current density If:
For a metal surface under a thin electrolytic film, when the A x B  x
thickness (d) of the electrolytic film is very small (d ? 0), the state IF ¼  exp  exp  ð3Þ
qp L qp L

A x B  x
If ¼  exp þ exp 
qs L L qs L L
A x B  x
¼  pffiffiffiffiffiffiffiffiffiffiffi exp þ pffiffiffiffiffiffiffiffiffiffiffi exp  ð4Þ
qs qp L qs qp L

where IF is a curve similar to E in shape, and If has a shape similar to


the differential of E (see Fig. 2). IF is in the direction perpendicular to
the metal/electrolyte interface whilst If is parallel to the interface. IF
flowing into the electrolyte and If flowing from left to right in this
study are defined to have a positive value.

2.2. ‘‘Dead end” and ‘‘open ends” systems

Based on the above general potential and current expressions, the


potential and current expressions for systems with controlled
boundary conditions can be obtained. In practice, this can be a case
where a piece of sheet or tube metal has a thin electrolyte film over
its surface, but the electrolyte or the metal is not infinite in length.
When a current or potential is applied through the electrolyte at
one end, the current If flowing in the electrolyte will always be zero
Fig. 1. Schematic illustration of two typical one-dimensional systems: (a) small at the other end (the ‘‘dead end”). If potentials or currents are applied
tube and (b) thin electrolyte coverage. at both ends, that will be a case of ‘‘open ends”.
458 G.-L. Song / Corrosion Science 52 (2010) 455–480

Fig. 2. Schematic illustration of a one-dimensional system and corresponding potential and current distributions.

For a piece of metal (c) having length c, if a potential or current is where


applied at the left end x = 0, then at the other boundary end x = c, sffiffiffiffiffiffi
which is a ‘‘dead end”, the flowing current density Icf should be zero: c qcp
 L ¼ ð12Þ
Icf ¼ Ic x¼c ¼ Icfc ¼ 0 ð5Þ qs

Moreover, the applied potential in the electrolyte is Wc0 at boundary where qcp is the polarization resistivity of metal (c). Wc0 is the poten-
x = 0. By substituting the boundary conditions into the general E, IF tials of the electrolyte in this system at the left end x = 0. Eccorr is the
and If expressions, the following expressions for this system can be corrosion potential of metal (c) relative to a reference electrode.
deduced (see Appendix C): Correspondingly, the Faradic and flowing current densities
  IcF and Icf over the metal can be obtained according to Eqs. (3) and
wc0 cosh xc
(4). Wcc is the potential of the electrolyte at the other end (x = c).
c c
w ¼  L ; 06x6c ð6Þ
cosh Lcc Icf0 is the current at boundary x = 0 where the boundary potential
 
wc cosh xc c is applied. The distributions of the potential and current densities
Ec ¼ Eccorr  0  c L ; 0 6 x 6 c ð7Þ are shown in Fig. 3.
cosh Lc
!   For a piece of metal (a) having a length ‘‘a”, if a current or poten-
c wc0 cosh xc c
tial is applied at its right end, then If = 0 at x = 0 which is a ‘‘dead
IF ¼  c L  ; 0 6 x 6 c ð8Þ
qp cosh Lcc end”. Similarly, the potential and current equations for this system
!   can be obtained:
wc0 sinh xc c
 x
Icf ¼  pffiffiffiffiffiffiffiffiffiffi
ffic L  ; 0 6 x 6 c ð9Þ
qs qp cosh Lcc x a cosh La
wa ¼ waa cosh ¼ w  ; 0 6 x 6 a ð13Þ
La a
cosh Laa
wc0  
wcc ¼ wc jx¼c ¼   ð10Þ wa cosh Lxa
cosh Lcc Ea ¼ Eacorr  a a ; 0 6 x 6 a ð14Þ
  cosh La
!  
 wc0 tanh Lcc a
wa cosh Lxa
Icf0 ¼ Icf x¼0 ¼ pffiffiffiffiffiffiffiffiffifficffi ð11Þ a
IF ¼  a  ; 0 6 x 6 a ð15Þ
qs qp qp cosh Laa
G.-L. Song / Corrosion Science 52 (2010) 455–480 459

Fig. 3. Schematic illustration of a one-dimensional system (c) with controlled boundary condition at its left end and corresponding current and potential distributions.

!  
sinh x If two boundary conditions Wb0 and Wbb are known, we can easily
waa La
Ifa ¼  pffiffiffiffiffiffiffiffiffiffi
ffi  ; 06x6a ð16Þ work out the potential and current distributions:
qs qp cosh
a a
La
 
 wa 
IaFa ¼ IaF x¼a ¼  aa ð17Þ wbb  wb0 exp  Lbb x
pq wb ¼   exp b
 b
2 sinh Lb L
 waa a
Iafa ¼ Iaf x¼a ¼  pffiffiffiffiffiffiffiffiffiffi
ffi tanh ð18Þ
qs qp
a La  
wb0 exp  Lbb  wbb 
x
where þ   exp  b ; 06x6b ð20Þ
2 sinh Lbb L
sffiffiffiffiffiffi
qap
La ¼ ð19Þ  
qs wbb  wb0 exp  Lbb 
b x
E ¼ Ebcorr
   exp b
where qap is the polarization resistivity of metal (a); Waa is the poten- 2 sinh Lbb L
tial of the electrolyte applied at boundary x = a. Wa0 is the potential  
wb0 exp Lbb  wbb 
of the electrolyte at the other end x = 0. Eacorr is the corrosion poten- x
þ   exp  b ; 0 6 x 6 b ð21Þ
tial relative to a reference electrode. The distribution of potential L
2 sinh bb L
and currents described by Eqs. (13)–(16) are shown in Fig. 4.
For a section of metal having length b (Fig. 5), if both of its 2 b    
b b  wb exp b  wb
boundary conditions are known, e.g. at the left end, the potential 1 4wb  w0 exp  Lb x 0 Lb b
b
of the electrolyte is Wb0 and current is i0 ; and at the right end, Ibf ¼  qffiffiffiffiffiffiffiffiffiffiffi   exp b   
b b qs qbp b
2 sinh Lb L 2 sinh Lbb
Wb and ib , respectively, then it is an ‘‘open ends” system. It should


be noted that only two of these boundary conditions are indepen- x


b b
dent. When Wb0 and Wbb are fixed, i0 and ib will be dependent on  exp  b ; ð22Þ
b b
L
W0 and Wb , and vice verse.
460 G.-L. Song / Corrosion Science 52 (2010) 455–480

Fig. 4. Schematic illustration of a one-dimensional system (a) with controlled boundary condition at its right end and corresponding current and potential distributions.

2 b  
b b  flowing current change largely at a boundary (or ‘‘open end”)
1 4wb  w0 exp  Lb x
IbF ¼  b   exp b where a boundary potential or current is applied. The changing
qp b
2 sinh Lb L amplitude decreases as the distance from the boundary in-
  3 creases. At a boundary (or ‘‘dead end”) where no potential or
w0 exp Lb  wbb
b b 
x 5 current is applied, the Faradic current and flowing current be-
þ   exp  b ; 0 6 x 6 b ð23Þ
2 sinh bb L come linearly dependent on distance. Moreover, it can be theo-
L
retically deduced that the applied overall polarization current
of a ‘‘dead end” or ‘‘open ends” system is a linear function of
where
the applied potential (see Appendix D), which is consistent with
sffiffiffiffiffiffi
the linear assumption in this study.
b qbp
L ¼ ð24Þ It should be noted that the above ‘‘dead end” and ‘‘open ends”
qs system can comprise a complicated galvanic corrosion system.
The potential and equation equations deduced for these elemental
where qbp is the polarization resistivity of the metal in section b, and
systems lay a foundation for deducting potential and current equa-
Ebcorr is the corrosion potential of material (b) relative to a reference
tions of complicated galvanic systems.
electrode.
The potential and current distributions in this system follow
exponential functions and are schematically shown in Fig. 5.
Potentials and current densities change more dramatically in a re- 3. Galvanic current, potential and their distributions
gion close to a boundary and less significantly farther away from a
boundary. If one of the above metals is in contact with other one or two
In summary, as long as the polarization currents or polariza- different metals, they will together form a galvanic system due
tion potentials at boundaries or ends are known, the potential to their different electrochemical activities. In this case, the exter-
and current distributions over a metal surface can be deter- nal polarization current becomes an internal current flowing be-
mined. The current equations deduced from the above systems tween two adjacent metals in connection. This flowing current is
suggested that both the Faradic current and the non-Faradic also known as a galvanic current of the system.
G.-L. Song / Corrosion Science 52 (2010) 455–480 461

Fig. 5. Schematic illustration of a section with controlled boundary conditions at both ends and corresponding current and potential distributions.

pffiffiffiffiffiffi   "  #
3.1. A joint of two dissimilar metals ðEacorr  Eccorr Þ qcp tanh Laa cosh xac
Lc 
Ec ¼ Eccorr þ hpffiffiffiffiffiffi   pffiffiffiffiffiffi   i  ;
qap tanh Lcc þ qcp tanh Laa cosh Lcc
If there are two pieces of metals (a) and (c) joined up (see Fig. 6),
each metal will be equivalent to a ‘‘dead end” single piece metal as a6x6aþc ð29Þ
a
discussed above, and the current ig flowing out of the electrolytic
c   "  #
film over metal (a) should be equal to the current ig into the elec- ðEccorr  Eacorr Þ tanh Lcc cosh Lxa
trolytic film over metal (c) across the point interface: IaF ¼ pffiffiffiffiffiffihpffiffiffiffiffiffi   pffiffiffiffiffiffi  i a ;
a ac c
qap qcp tanh Laa þ qap tanh Lcc cosh La
ig ¼ ig ¼ ig ¼ ip ð25Þ
06x6a ð30Þ
or
  "  #
a c
ig  ig ¼ 0 ð26Þ ðEa  Eccorr Þ tanh Laa cosh xac
La 
IcF ¼ pffiffiffiffiffiffihpffiffiffiffifficorr
ffi   pffiffiffiffiffiffi   i  ;
qcp qap tanh Lcc þ qcp tanh Laa cosh Lcc
Meanwhile, for these two pieces of metals, the potential Ea of metal
(a) should be equal to the potential Ec of metal (c) at their joint a6x6a ð31Þ
interface:
    "  #
Ea x¼a ¼ Ec x¼0 ð27Þ ðEccorr  Eacorr Þ tanh Lcc sinh Lxa
Iaf ¼ pffiffiffiffiffihpffiffiffiffiffiffi   pffiffiffiffiffiffi  i a ;
With the above relationships, the potential and current of the sys-
qs qcp tanh Laa þ qap tanh Lcc cosh La
tem can be obtained (see Appendix E): 06x6a ð32Þ
pffiffiffiffiffiffi   "  #   "  #
a a
ðEccorr  Eacorr Þ qap tanh Lcc cosh Lxa ðEa  Ec Þ tanh Lac sinh xac
E ¼ Ecorr þ hpffiffiffiffiffiffi Lc 
  pffiffiffiffiffiffi  i a ; Icf ¼ pffiffiffiffiffihpffiffiffiffiffifficorr corr
 pffiffiffiffiffiffi   i  ;
qc tanh aa þ qa tanh cc cosh La
p L p L qs qap tanh Lcc þ qcp tanh Laa cosh Lcc

06x6a ð28Þ a6x6aþc ð33Þ


462 G.-L. Song / Corrosion Science 52 (2010) 455–480

Fig. 6. Schematic illustration of two coupled systems (a) and (b) and corresponding current and potential distributions.

ac
  ðEccorr  Eacorr Þ ‘‘dead end”, and material (b) that is inserted between metals (a)
ig ¼ Iaf x¼a ¼ Icf x¼a pffiffiffiffiffihpffiffiffiffiffiffi.   pffiffiffiffiffiffi.  i ð34Þ
qs qap tanh Laa þ qcp tanh Lcc and (c) can be treated as a sectional system with ‘‘open ends”.
Similarly, the potentials and flowing currents over different
metals must be equal at their joints. These will lead to their theo-
 ðEccorr  Eacorr Þ
IaF0 ¼ IaF x¼a ¼ pffiffiffiffiffiffihpffiffiffiffiffiffi  .   pffiffiffiffiffiffii ð35Þ retical potential and current expressions (see Appendix F):
qap qcp tanh Laa tanh Lcc þ qap " #  
MEbc þ HEba cosh Lxa
IaF ¼   Eba a  ; 06x6a ð37Þ
 ðEacorr  Eccorr Þ GH  MN qp cosh Laa
IcF0 ¼ IcF x¼a ¼ pffiffiffiffiffiffihpffiffiffiffiffiffi  .   pffiffiffiffiffiffii ð36Þ
qp
c qp tanh Lcc tanh Laa þ qcp
a
 
NEba þGEbc bc
þHEba
 MEGHMN exp  Lbb 
The potential and current distributions of this system are schemat- GHMN xa
IbF ¼   exp
ically displayed in Fig. 6. Basically, the potential and current curves Lb
2qbp sinh Lbb
over these two joined metals have the same characteristics as those  
before they are joined together. Furthermore, their Faradic currents MEbc þHEba ba
exp Lbb  NEGHMN þGEbc 
GHMN xa
are always negative to each other according to equations of    exp  b ;
IaF and IcF . This suggests that these two metals are always anodic to 2qbp sinh Lbb L
cathodic each other. a6x6aþb ð38Þ

3.2. Joints of three pieces of metals in series !  


NEba þ GEbc cosh xabc
Lc 
If there are three pieces of materials (a), (b) and (c) joined in ser- IcF ¼   Ebc ;
GH  MN qcp cosh Lcc
ies as such that (a) is joined to (b) which is joined to (c) (see Fig. 7),
then materials (a) and (c) will be equivalent to a system having a aþb6x6aþbþc ð39Þ
G.-L. Song / Corrosion Science 52 (2010) 455–480 463

Fig. 7. Schematic illustration of a thin electrolytic film covering metal (a) joined to metal (b) joint joined to metal (c) and corresponding current and potential distributions.

" #   !  
MEbc þ HEba sinh Lxa  NEba þ GEbc tanh Lcc
Icf x¼aþb
bc
IaF ¼ ba
 E pffiffiffiffiffiffiffiffiffiffiaffi  ; 06x6a ð40Þ ig ¼ Icf0 ¼ ¼  Ebc pffiffiffiffiffiffiffiffiffifficffi ð44Þ
GH  MN qs qp cosh Laa GH  MN qs qp
  " #
NEba þGEbc bc
 MEGHMNþHEba
exp  Lbb   1 MEbc þ HEba
GHMN xa IaFa ¼ IaF x¼a ¼  a  Eba
Ibf ¼ qffiffiffiffiffiffiffiffiffiffiffi   exp qp GH  MN
ð45Þ
2 qs qbp sinh Lbb Lb
  !
MEbc þHEba
exp Lbb  NEGHMN
ba
þGEbc  
GHMN xa  1 MEbc þ HEba
þ qffiffiffiffiffiffiffiffiffiffiffi   exp  b ; IbF0 ¼ IbF  ¼ ð46Þ
2 qs qbp sinh Lbb L x¼a qbp GH  MN

a6x6aþb ð41Þ !
 1 NEba þ GEbc

!   IbFb ¼ IbF  ¼ ð47Þ
NEba þ GEbc sinh xabc c
x¼aþb qbp GH  MN
Icf ¼   Ebc pffiffiffiffiffiffiffiffiffifficffi L  c  ; aþb 6x
GH  MN qs qp cosh Lc !
 1 NEba þ GEbc
6aþbþc ð42Þ IcF0 ¼ IcF x¼aþb ¼  c  Ebc ð48Þ
qp GH  MN
" #  
ab
 MEbc þ HEba tanh Laa
ig ¼ Iaf x¼a ¼   Eba pffiffiffiffiffiffiffiffiffiffi
ffi ð43Þ where M, N, G and H are functions of systematic parameters of the
GH  MN qs qap three materials as defined in Appendix F. Fig. 7 schematically
464 G.-L. Song / Corrosion Science 52 (2010) 455–480

presents the potentials and currents of this system. All the three Parameters mi, Li, qs and qi are constants that are known or can be
connected materials can actually affect one another in their poten- measured, so in total 2n variables wi0 and wib are involved in Eq. (53).
tial and current distributions. For example, currents over material They are all simple linear equations. The number of equations is n.
(a) are not only influenced by material (b) that is in direct connec- At the same time, we know:
tion with it, but also dependent on the parameters of material (c)
which is not directly in contact with metal (a). This is termed as a W00  W10 ¼ E0corr  E1corr
‘‘remote” effect in this study.
W1b  W20 ¼ E1corr  E2corr
3.3. A complicated one-dimensional system 
i1
Wbi1  Wi0 ¼ Ecorr  Eicorr ð54Þ
The methodology adopted to deduce the potentials and currents

for the galvanic systems with two joints and three pieces of mate-
n2
rials is straightforward; simply let the flowing currents and poten- Wn2
b  Wn1
0 ¼ Ecorr  En1
corr
tials of the two materials in direct connection be equal at their Wn1  Wn0 ¼ En1 n
b corr  Ecorr
joint point. This will generate a sufficient number of equations to
determine the unknown parameters of these galvanic systems. In
Equation set (54) provides additionally n linear equations. There-
fact, the methodology can be extended to work out potentials
fore, all the potentials wi0 and wib can be worked out and have their
and current densities for systems containing more than three
analytical expressions. Correspondingly, the galvanic potentials and
pieces of materials joined together in series. For a multiple-mate-
currents over each piece of the element can be obtained. In other
rial system, in addition to the left and right ‘‘dead end” pieces m0
words, the galvanic potential and current distributions over a mul-
and mn, there are n-1 ‘‘open ends” pieces between m0 and mn
ti-piece system will have analytical expressions. Although the
(see Fig. 8). In this case, the polarization currents flowing in/out
expressions are more complicated than those for a three-piece sys-
these ‘‘dead end” and ‘‘open ends” pieces still following Eqs. (11),
tem, it is a great advantage to analytically calculate a theoretical
(18), (97) and (98):
  galvanic potential or current directly based on system parameters,

 w0b tanh mL00 rather than to simulate them through a numerical modeling
0
ib ¼ I0f  ¼ qffiffiffiffiffiffiffiffiffiffiffi ð49Þ process.
x¼0
qs q0p It should be noted that this multi-piece system can be easily ex-
 tended to a more complicated non-uniform system. For example, a
 wn0 mn
in ¼ Inf  P
0 non-uniform system can be divided into many small pieces and
n1 ¼  pffiffiffiffiffiffiffiffiffiffi
ffin tanh ð50Þ
x¼ mi qs qp La each of the small pieces can be regarded as a uniform system.
i¼0
  Therefore, the complicated non-uniform system becomes a mul-
mi

 wi0 cosh Li
 wib ti-piece system.
i
i0 ¼ Iif  P
i1 ¼ qffiffiffiffiffiffiffiffiffiffiffi   ð51Þ Due to the complexity of potential and current expressions of a
mi
x¼ mj qs qip sinh Lb multi-system, in the following sections, only 2 and 3 piece joint
ji¼0
  systems are discussed. However, the conclusions hold for a compli-

 wi0  wib cosh mLii cated system as well.
i
ib ¼ Ibf  P i ¼ qffiffiffiffiffiffiffiffiffiffiffi   ð52Þ
x¼ mj qs qip sinh mLii
ji¼0
4. Parameters and critical values
They should be equal between every two adjacent pieces:
The above deduced galvanic potential and current equations are
0 1
ib ¼ i0 actually determined by many parameters, such as the polarization
1 2 resistivity, solution resistivity, length of materials, etc. Among
ib ¼ i0 these parameters, some can affect the characteristics of potential
 and current distributions more significantly than the others.
i1 i
ib ¼ i0 ð53Þ
4.1. System determination parameters

n2 n1
ib ¼ i0 A galvanic system has three different types of parameters: (1)
n1 n system geometric parameters, such as the lengths of systems a, b
ib ¼ in
and c; (2) material parameters, Eacorr ; Ebcorr ; Eccorr ; qap ; qbp ; qcp , and

Fig. 8. Schematic illustration of a multi-piece system (complicated system).


G.-L. Song / Corrosion Science 52 (2010) 455–480 465

sffiffiffiffiffiffi sffiffiffiffiffiffiffiffi sffiffiffiffiffiffiffiffi


(3) environmental electrolyte parameter qs. It should be noted that qp q0p S q0p d
in this study, qs is not a pure solution property. It also contains L¼ ¼ ¼ ð55Þ
qs qs y
0 q0s
geometric parameters y, r and/or d (Appendix A).
According to the above general potential and current equations,
the influence of these parameters on galvanic potential and current which means that L is actually a combination of the traditional
distributions can be summarized as follows. The geometric param- ‘‘Wagner Number” and the electrolyte film thickness. L can be trea-
eters only affect the amplitudes of galvanic potentials and currents. ted as a one-dimensional ‘‘Wagner Number”. It has also a unit of
Some of the material parameters, such as the corrosion potentials, distance, and can be regarded as a measure of the effective distribu-
will determine the amplitude of galvanic potentials and currents, tion distance within which significant galvanic current or potential
while others, such as qap ; qbp and qcp , will have an influence on po- is distributed on the metal surface [29]. So, it is termed as ‘‘effective
tential and current distributions. The environmental electrolytic distribution distance” in this study. If a galvanic current distribution
parameter qs can significant affect the potential and current distri- is concerned, L is actually the distance from a couple joint to a point
butions as well as their amplitudes. Generally speaking, an in- where the slope of the galvanic current distribution curve at the
crease in the difference between the corrosion potentials of the junction point intersects the x-coordinate (see Figs. 2–4). Within
metals will lead to an increase in galvanic current density, but an this distance, the galvanic current or potential of a system decreases
increase in electrolytic resistivity or polarization resistivity of the most dramatically, and after that, the change becomes steady.
metals will result in a decrease in galvanic currents. It should be For an infinite system, L is a useful measure for the distance that
stressed that the dependence of galvanic potentials and currents a galvanic potential or current is distributed. However, in a system
on the parameters is not straightforward. Due to an interaction be- with a limited length, this measure becomes invalid, particularly
tween the metals, sometimes the influence is complicated, partic- when the length of the system is shorter than L. For example, in
ularly for a system with more than two pieces of materials Fig. 9, when Lc > c, exponential distribution of IcF cannot be well dis-
connected in series. played in the entire cathode surface. There is a possibility that two
galvanic systems can have the same L due to their similar qs and
qp, but their current and potential distributions may appears to
4.2. Distribution parameter be different because of their different material length limitations.
An example is given in Fig. 10.
Among those parameters determining the potentials and cur- In fact, variable x in the above potential and current expressions
rents of a galvanic system, there are two special parameters that is always restricted within the length of a corresponding metal.
have a critical influence on the potential and current distributions: According to the galvanic potential and current expression in Sec-
(1) the material polarization resistivity qp and (2) the environmen- tion 3, a potential or current distribution determined by exponen-
tal electrolyte resistivity qs. L is a square root of their ratio. A sys- tial functions is eventually determined by the ratio of x/L, not by x
tem with a larger L will have a potential and current changing or L alone. It is quite possible in mathematics that two galvanic
relatively less significantly with distance. The influence of the L va- systems which have different L values show similar potential and
lue on the current distributions of a galvanic system is clearly illus- current distribution curves in different x scales. To eliminate the
trated in Fig. 9. influence of material length on a potential or current distribution
In fact, the ratio of polarization resistivity over solution resistiv- curve, x should be transformed into a non-dimensional variable
ity has been termed as ‘‘Wagner Number” and used to predict the X. Let
galvanic current distribution of a 3-D infinite galvanic system
x
[11,27,28]. ‘‘Wagner Number” has a unit of distance. The only dif- x
ference between ‘‘Wagner Number” and parameter L defined in ¼ mL  ¼ WX ð56Þ
L m
this study is the square root involved in L. This is because ‘‘Wagner
Number” and L are defined in different dimensional systems.
According to Eqs. (77) and (79), we have: then

Fig. 9. Theoretical Faradic and flowing currents (IF and If) over an anode (a)– Fig. 10. Theoretical Faradic and flowing currents (IF and If) over an anode (a)–
cathode (c) couple with La  Lc. The curves are produced using corresponding cathode (c) couple with length a  length c. The curves are produced using
equations in Section 3. corresponding equations in Section 3.
466 G.-L. Song / Corrosion Science 52 (2010) 455–480

x A combination of the above three cases can also lead to a very


X¼ ð57Þ
m large W value.
L
W¼ ð58Þ
m 4.2.2. W ? 1
where m is material length, like a, b and c; L can be La, Lb and Lc; and W can be very large at least in the following three cases: (1)
W is a new distribution parameter, which can be Wa, Wb and Wc for large m (e.g. a, b or c); (2) large qs; and (3) small qp (e.g.
materials (a), (b) and (c). qap ; qbp ; qcp ).
After the transformation, X and W become dimensionless. In
mathematics, X will always range from 0 to 1 and W is purely a curve (1) Large m (e.g. a, b or c) or m ? 1: This can be a case when the
‘‘distribution characteristic number”, which specifies the depen- material length of a studied system is much larger than the
dence of galvanic potential or current on distance x. W is a combina- effective distribution distance.
tion of geometric effect, material effect and electrolytic effect (L only (2) Large qs or qs ? 1: This can be a system in some electro-
represents the effect of material and environmental properties). A lytes, such as extremely diluted solutions, tap water, rain
large W means that potentials or currents vary significantly with X, water, condensed water, pure water, some special coolants,
whereas they change slowly with X if W is small. On a transformed fuels, brake fluids, anti-corrosion oils and detergents. Some-
X-coordinate, different systems should have a similar potential or times, such a system may not be noticed. For example, an
current curve if their ‘‘distribution characteristic numbers” are the organic coating that has taken up sufficient water after being
same. Fig. 11 shows the Faradic and flowing currents (IF and If) over exposed in service environments for a long time. While a
an anode (a)–cathode (c) couple with different La and Lc but the same metal coated with such a coating is in a dry environment,
Wa and Wc. The currents appear to be mainly distributed within the high resistivity coating will act as an electrolytic film
about 0.3 of the lengths over the anode and cathode from their joint for the coated metal.
point, even through the anode and cathode have very different (3) Small qp (e.g. qap ; qbp ; qcp ) or qp ? 0: Some active metal sur-
lengths, polarization resistivity, and L values. faces without any surface protection can have a very fast
According to the value of ‘‘distribution characteristic number” Faradic reaction. Sometimes, a passive metal after transpass-
W, galvanic corrosion systems can be easily classified into different ivation, reactivation or suffering from pitting corrosion or
groups. crevice corrosion can also have very small polarization
resistivity.
4.2.1. W ? 0
There are at least three possibilities that parameter W can be A combination of the above three cases can also lead to a very
very small: small W value.

(1) Small m (e.g. a, b or c) or m ? 0: This can be a case that the 4.3. Simplified potential and current equations in some extreme cases
material length of a studied system is much smaller than the
effective distribution distance. When W ? 0 or 1, the theoretical potential and current expres-
(2) Small qs or qs ? 0: This can be an electrolyte having very sions in Section 3 can be simplified (see Appendix H). In these
good conductivity, such as concentrated salt solution, sea cases, the potential and current curves will have some characteris-
water, deicing salt mixed with melted snow, concentrated tics that can be utilized to estimate the corrosion damage of a gal-
acidic or alkaline solutions. vanic system in practice.
(3) Large qp (e.g. qap ; qbp ; qcp ) or qp ? 1: This refers to a difficult For a system with small W, the potential and current expressions
Faradic reaction. In theory, the polarization resistivity will become linear equations according to Appendix H.1. The curves in
be very large when the material is not active, or in a passive Figs. 12 and 13 are produced using corresponding equations in
state with a stable passive film protection, or coated with an Section 3. They show that flowing current densities Iaf ; Ibf and Icf are
inert corrosion resistant film, or the Faradic reaction at the linearly dependent on distance x in the system with large distribu-
metal surface is controlled by a diffusion step. tion numbers Wa, Wb and Wc. This is in agreement with the linear dis-
tribution of flowing current densities as predicted by those
simplified equations in Appendices H.1.1 and H.1.2.
For a system of two joined dissimilar metals with small Wa and
c
W , according to Appendix H.1.1 the portion of the potential differ-
ence distributed over a metal is proportional to the polarization
resistance of this metal. The flowing current densities in the elec-
trolyte are linearly dependent on the distance. In addition to the
corrosion potential difference of these two metals, the sum of
polarization resistances of these two metals also governs the flow-
ing current and Faradic reaction current. The currents are recipro-
cally proportional to the material length and the total polarization
resistance. These potentials and currents become independent
from the resistivity of the electrolyte in this case due to small Wa
and Wc values. The galvanic current between these two metals be-
come proportional to both the corrosion potential difference and
the reciprocal of the sum of polarization resistances. This is consis-
tent with the simple conventional galvanic equation that is nor-
mally used to predict galvanic current between two metals [29].
Similarly, the maximum galvanic current densities of metals (a)
Fig. 11. Theoretical Faradic and flowing currents (IF and If) over an anode (a) and
and (c) at the joint are also simply determined by the corrosion po-
cathode (c) couple with different La and Lc but the same Wa and Wc. The curves are tential difference and the sums of polarization resistances and
produced using corresponding equations in Section 3. lengths of these two metals. A larger corrosion potential difference
G.-L. Song / Corrosion Science 52 (2010) 455–480 467

Fig. 13. Theoretical potentials and currents (IF and If) over a system of three pieces
in series with small Wa, Wb and Wc. The curves are produced using corresponding
Fig. 12. Theoretical potentials and currents (IF and If) over an anode (a)–cathode (c) equations in Section 3.
couple with small Wa and Wc. The curves are produced using corresponding
equations in Section 3.
asymmetric (referring to Appendix H.3), which has been shown
in Fig. 9.
When a small W material is inserted between two other materi-
and smaller polarization resistance of the metals will result in a als, an interesting prediction about galvanic current densities can
higher maximum galvanic current density. Increasing the length be made according to Appendix H.4. The galvanic currents at joints
of a metal will decrease its maximum current density. ‘‘a/b” and ‘‘b/c” are equal and independent from the corrosion poten-
If the two or three dissimilar metals joined in series have large tial of metal (b). In fact, Wb ? 0 means b ? 0 or qbp ! 1 or qs ? 0. In
W values, then the galvanic potentials and currents will have any of these cases, the flowing galvanic current consumed by the
sinh(x/L) and cosh(x/L) distributions (see Appendix H.2 or H.2.1 Faradic reaction over metal (b) can be neglected, and all the galvanic
and Appendix H.2.2). In the region next to a joint, they can be fur- current flowing into metal (b) from joint ‘‘a/b” will flow out from
ther simplified into exponential distributions. In fact, the poten- joint ‘‘b/c”. Hence, the galvanic currents at these two joints are nearly
tials and currents vary significantly within a narrow region equal. Since the total Faradic current consumed by the system is neg-
adjacent to the joints. The galvanic potentials and currents will be- ligible, the galvanic potential will have nothing to do with the total
come smaller in regions far away from the joints. Figs. 14 and 15 galvanic current. For the same reason, the Faradic current or the cor-
show the galvanic potentials and currents of a system with large rosion potential of metal (b) will have a very limited influence on the
W values, in which the currents are nearly zero and potentials potential and current distributions over metals (a) and (b). However,
are constant in most regions except that they vary dramatically the corrosion potential of metal (b) can significantly affect the poten-
within a very narrow region next to the joints. Moreover, according tial and currents over metal (b). Fig. 16 shows the currents and
to Appendices H.2.1 and H.2.2, the galvanic current between two potentials of such a system.
adjacent metals are dependent on the polarization resistivity of In case that a large W material is insert between two small W
both metals and the conductivity of the electrolyte, while the max- materials (referring to Appendix H.5), the galvanic current at the
imum galvanic current density at the joint is purely determined by connection points will be not only dependent on the potential dif-
the polarization resistivity of the adjacent metals and the corrosion ference of the adjacent metals, but also on their polarization resis-
potential difference, but independent from electrolyte film tivity and the electrolyte resistivity. The maximum galvanic
conductivity. current densities are also mainly determined by the adjacent mate-
If a piece of large W metal is in contact with a small W metal, rial parameters. This is because the large W metal between the two
then the galvanic potentials and currents over the couple will be small W materials effectively consumes the currents from the
468 G.-L. Song / Corrosion Science 52 (2010) 455–480

Fig. 15. Theoretical potentials and currents (IF and If) over a system of three pieces
Fig. 14. Theoretical potentials and currents (IF and If) over an anode (a)–cathode (c) in series with large Wa, Wb and Wc. The curves are produced using corresponding
couple with large Wa and Wc. The curves are produced using corresponding equations in Section 3.
equations in Section 3.

5. Case studies
small W materials, and thus the separated small W metals have no
interaction. Corresponding potential and current distributions of The developed galvanic potential and galvanic current equa-
this system are shown in Fig. 17. tions are useful tools in analyzing galvanic corrosion. Many practi-
cal cases can have a suitable theoretical system as classified above
to present, and their galvanic potential and current distributions
4.4. Galvanic current characterization parameters can be easily predicted by using the corresponding theoretical
equations. In this section, some examples are provided to show
In practice, engineers are concerned about two aspects of gal- how the theoretical equations can be utilized to address specific
vanic damage to a system: (1) the overall damage (quantified by a to- galvanic corrosion issues.
tal galvanic current) and (2) the damage intensity (quantified by a
maximum galvanic current density). The total galvanic current and
the maximum galvanic current density also characterize a galvanic 5.1. Experimentally measured galvanic current densities
current distribution curve or the galvanic effect of a system, and thus
they can be taken as two important galvanic effect characterization A comparison between theoretically predicted and experimen-
parameters. Obviously the total galvanic current ig can be easily ob- tally measured galvanic current densities is presented in Fig. 18.
tained according to the equations deduced earlier. A maximum Fara- The experimentally measured points are obtained from reference
dic current density only occurs at a joint of a galvanic system. [29]. In that reference, the galvanic current densities are denoted
Therefore, Im m
F0 or I Fm for material (m) (m can be a, b or c) with a length as Ig, but they are actually IF according to the definition in this pa-
of ‘‘m” represents a maximum current density. In this study, galvanic per. It should be noted that the anodic polarization curve of Mg
effect characterization parameters refer to ig, Im m
F0 and I Fm . (AZ91D) is nearly linear while none of the other three metals
It should be stressed that both a positive (anodic) and negative (Zn, Al and steel) has a perfect linear catholic polarization curve
(cathodic) Faradic currents can have a damaging effect on a cou- in 5 wt% NaCl solution [29]. As the catholic polarization curve of
pling metal. A positive Faradic current certainly results in galvanic Zn is relatively better than Al and steel in terms of linearity, the
corrosion damage to the metal, while a significantly large negative experimentally measured current densities of the Mg–Zn couple
Faradic current density can give rise to an over-protection of a me- are used in this study to compare with theoretically predicted gal-
tal, which is also a type of damage. vanic current densities.
G.-L. Song / Corrosion Science 52 (2010) 455–480 469

Fig. 17. Theoretical potentials and currents (IF and If) over a system of three pieces
Fig. 16. Theoretical potentials and currents (IF and If) over a system of three pieces in series with small Wa and Wc but large Wb. The curves are produced using
in series with large Wa and Wc but small Wb. The curves are produced using corresponding equations in Section 3.
corresponding equations in Section 3.

where galvanic current densities are higher, whereas in a region


In Fig. 18 the theoretically predicted anodic and catholic gal- with lower current densities far away from the joint it becomes lar-
vanic current density curves (the solid lines) are calculated accord- ger. Such a non-constant qp over a metal is contradictory to the
ing to the galvanic current density expression in Section 3 using assumption and can result in a theoretically predicted galvanic
some parameters estimated from the polarization curve of Mg current density different from its corresponding experimental
(AZ91D) immersed in the same 5 wt% solution in an electrolyte cell point.
[29]. It appears that there is some reasonable agreement between In this study, the relatively significant discrepancy over the Zn
the predicted curves (solid lines) and the experimental points. At surface than over the Mg surface may be ascribed to the worse lin-
least, it is correctly predicted that the anodic and cathodic galvanic earity of the catholic polarization curve of Zn than the anodic
current densities are asymmetrically distributed over the Mg and polarization curve of Mg (AZ91D). The anodic polarization curve
Zn surfaces; the anodic galvanic current densities in the region of Mg is nearly linear and thus its qap can be reasonable assumed
next to the MgjZn joint are much higher but decrease much faster to be constant (independent from distance) over the Mg surface.
with the distance from the joint on the Mg surface than the cath- For Zn, its catholic polarization curve is obviously non-linear. In
olic galvanic current densities on the Zn surface (see the points this study, only a small section (close to be linear) of its catholic
and solid curves in Fig. 18). polarization curve is used to estimate qcp and Eccorr . The estimated
However, evident deviation of the experimental points from the qcp and Eccorr may only represent its polarization resistivity and cor-
theoretical curves (solid lines) is also seen in Fig. 18. This is under- rosion potential in a particular surface area. In other surface areas,
standable. The analytical equations deduced in previous sections they may actually be significantly different from those estimated
are based on some assumptions, such as constant polarization ones. Therefore, if these estimated qcp and Eccorr are used to calcu-
resistivity qp, constant solution resistivity qs and stable corrosion late the theoretical galvanic current curves of the Mg–Zn couple,
potential Ecorr. In practice these assumption requirements cannot a significant difference will result between predicted and mea-
be strictly met. sured results.
The constant polarization resistivity assumption requires that It is also required by the constant qs assumption that the elec-
the qp of a metal does not change with distance, which means that trolyte film has a uniform composition, concentration and thick-
at least the polarization curve of the metal is linear. Unfortunately, ness over the entire galvanic couple. However, due to a non-
a practical metal cannot have a strictly linear the polarization uniform surface tension force over the surface in practice, the elec-
curve over a wide potential range. Quite often, qp is current density trolyte film may be much thin or disappear in some surface areas.
dependent. It is usually smaller in a region next to the couple joint Also, the Faradic reactions on the cathode and anode surfaces can
470 G.-L. Song / Corrosion Science 52 (2010) 455–480

Fig. 18. Theoretically predicted distribution of galvanic current density and experimentally measured galvanic current densities of a Mg–Zn galvanic couple under the ASTM
B-117 standard salt spray condition. The experimental points are from Ref. [29]. *qcp and Ecorr are estimated from the linear region of the catholic polarization curve of Zn
immersed in 5 wt% solution [29]. **Trial values assuming that the qcp and Ecorr become smaller and more positive under a salt spray condition than under an immersion
condition.

cause a change in composition and concentration of the electrolyte dicted curves (solid lines) deviate from the experimentally mea-
film. Particularly in the region adjacent to the anodejcathode joint, sured points on both Mg and Zn surfaces. It is believed that the
the corrosion products from the anode and cathode may interact theoretical prediction can be significantly improved if more accu-
each other and lead to a more complicated change in the electro- rate and reasonable parameters qp, qs and Ecorr are used. This is
lyte. The variation in electrolyte thickness only affects qp. A change demonstrated in Fig. 18. The dash curves in the figure are theoret-
in composition or concentration of the electrolyte will influence ically predicted galvanic current densities using a smaller trial qcp
not only qp but also qs. It has been demonstrated that sometimes and a more positive trial Eccorr based on the fact that sufficient oxy-
there is a ‘‘short-cut” effect on a galvanic couple, which is a result gen supply can accelerate the cathodic process of Zn underneath a
of non-uniform distribution of the electrolyte over a galvanic cor- thin electrolyte film under a salt spray condition. One can see that
rosion couple under a salt spray condition [29]. Therefore, it is be- the dash curves match the experimental points much better than
lieved that violation of the constant qs assumption should also be the solid curves on both Zn and Mg sides of the Mg–Zn couple.
one of the causes of the discrepancy between the experimental
points and the theoretical curves (solid lines) of the Mg–Zn couple
(Fig. 18). 5.2. ‘‘Linear superposition”
In addition to the above errors introduced by unsatisfactory
assumption requirements, the different conditions for measuring It was reported that in a ‘‘steel–AZ91D–steel” galvanic system
the galvanic current density over the AZ91D is equal to the sum
the polarization curves and for measuring the galvanic current
densities could also contribute to the discrepancy between the or addition of the galvanic current densities of AZ91D in two sep-
arated ‘‘steel–AZ91D” and ‘‘AZ91D–steel” systems based on com-
experimental points and theoretical curves. The galvanic current
densities were measured under a salt spray condition, whereas puter numerical analysis [24]. The phenomenon was termed as
‘‘linear superposition” in that paper. However, strictly speaking,
the polarization curves were obtained under an immersion condi-
tion [29]. One of the most important differences between immer- there is a possibility that the current density of AZ91D in the
‘‘steel–AZ91D–steel” system were accidently close to the addition
sion and spray is the supply of oxygen, which can significantly
affect a cathodic polarization curve of a conventional metallic or sum of the current densities of AZ91D in the ‘‘steel–AZ91D”
and ‘‘AZ91D–steel” systems, and simply because that the differ-
material. It is believed that the real catholic polarization curve of
ences between the current density and the sum are not significant,
Zn should have much higher cathodic current densities under a salt
they were mistaken as being equal. Now with the theoretical Fara-
spray condition than under an immersion condition. In other
dic current density expression in this study, the so-called ‘‘linear
words, the qcp and Eccorr of Zn during the galvanic current measure-
superposition” phenomenon can be analytically examined.
ment in a salt spray chamber should be much larger and more po-
We can assume that a = c, La = Lc, Wa = Wc, where materials (a)
sitive, respectively, than those estimated from its polarization
and (c) refer to steel and material (b) represents AZ91D. The origin
curve measured in an electrolyte cell. Evidently smaller theoretical
point of x-coordinate is set at the left end of the ‘‘steel–AZ91D–
galvanic current densities will be predicted if the qcp and Eccorr esti-
steel” system. For the couple of ‘‘steel–AZ91D” on the left in the
mated from the polarization curve of Zn are used in this case,
system, the galvanic current density over AZ91D can be expressed
which is exactly what are presented in Fig. 18 (the solid lines lower
as follows according to Eq. (31):
than the points). It should be noted that the influence of oxygen on
anodic dissolution of Mg is insignificant [30–32] and hence the   2  3
qap and Eacorr of Mg (AZ91D) estimated from its polarization curve Eacorr  Ebcorr tanhðW a Þ cosh xab
L b
b=a
IF ¼ pffiffiffiffi hpffiffiffiffiffiffi qffiffiffiffiffiffi i4 5;
in the electrolyte cell should be reasonable for theoretical predic- coshðW b Þ
qbp qap tanhðW b Þ þ qbp tanhðW a Þ
tion of the galvanic current densities of the Mg–Zn couple in the
salt spray chamber. Nevertheless, the inaccurate qcp and Eccorr can a6x6aþb ð59Þ
actually influence the prediction of the galvanic current densities
not only over Zn, but also on the Mg (AZ91D) surface according For the couple ‘‘AZ91D–steel” on the right in the system, x-coordi-
to the discussion in Section 3. Therefore, the theoretically pre- nate should be right shifted for a distance of ‘‘a” to quote Eq. (30):
G.-L. Song / Corrosion Science 52 (2010) 455–480 471

  2  3
Eccorr  Ebcorr tanhðW c Þ cosh xa may be explained according to the theoretical galvanic corrosion
Lb 5
Ib=c
F ¼ pffiffiffiffi hpffiffiffiffiffiffi qffiffiffiffiffiffi i4 b
; models developed above.
b b c
qp qp tanhðW Þ þ qp tanhðW Þ coshðW Þ
c b
The fuel and combustion systems of a car includes various pipe-
lines and connections. They can be treated as a one-dimensional
a6x6aþb ð60Þ
system. When fuel is taken in through the valves prior to combus-
The sum or superposition of these two current densities is: tion in the cylinder bores of a engine block, there should be a very
  h    i thin liquid fuel deposited on the surface of the valves. Therefore,
Eacorr  Ebcorr tanhðW a Þ cosh xab
Lb
þ cosh xa
Lb
around the in-take valve, a one-dimensional system is formed.
Ib=a b=c
F þ IF ¼ pffiffiffiffi hpffiffiffiffiffiffi qffiffiffiffiffiffi i ð61Þ E85 has conductivity around the order of lS/cm (MX cm) [33].
b
qp qap tanhðW Þ þ qbp tanhðW Þ coshðW b Þ
b a
For gasoline with 0% addition of ethanol is normally considered
as an insulator which has very low conductivity (can be as low
This is a symmetric galvanic system, in which
as 108 lS/cm). Therefore, the corroding area in a fuel system or
Eacorr ¼ Eccorr ; Eba ¼ Ebc ; Wb0 ¼ Wbb ; M ¼ N; and G ¼ H. Hence, Eq. (38)
engine block head can be considered as a large W galvanic system.
can be simplified:
For simplicity, if only a connection with two different metals is
  n    o
Eacorr  Ebcorr tanhðW a Þ sinh xa  sinh xab considered, then Eq. (184) (referring to Appendix H.2.1) can be
b b
L L
IbF ¼  pffiffiffiffi npffiffiffiffi pffiffiffiffiffiffi  b o ; used:
qbp qbp sinðW b Þ tanhðW a Þ þ qap sinh2 W2  c 
ac E  Eacorr
a6x6aþb ð62Þ ig  pffiffiffiffiffihcorr
pffiffiffiffiffiffi pffiffiffiffiffiffii ð65Þ
qs qap þ qcp
Thus,
The equation indicates that the overall galvanic corrosion damage is
Ib=a b=c
F þIF determined by the corrosion potentials, the polarization resistivity
IbF of the connected metals and the electrolyte resistivity.
h    inqffiffiffiffiffiffi pffiffiffiffiffiffi  b o
cosh xab þcosh xb qbp sinhðW b ÞtanhðW a Þþ qap sinh2 W2 There is a lack of information about the polarization resistivity
Lb Lb
¼n    ohpffiffiffiffiffiffi qffiffiffiffiffiffi i ð63Þ and the corrosion potentials of metals in E85. Generally speaking,
sinh xa sinh xab qap tanhðW b Þþ qbp tanhðW a Þ coshðW b Þ
Lb Lb the difference in corrosion potential of two different metals cannot
exceed 5 V. Hence, the influence of the corrosion potentials on the
Obviously,
overall galvanic corrosion can be assumed to be insignificant in this
Ib=a b=c
F þ IF
case. Al can react with ethanol fuel [34], implying that its polariza-
–1 ð64Þ tion resistivity may be lower in ethanol than in gasoline. As men-
IbF
  tioned above, the resistivity of E85 is more than eight orders of
i.e., Ib=a
F þ Ib=a
F cannot always be equal to IbF . In other words, the magnitude lower than that of gasoline. According to Eq. (184), this
‘‘linear superposition” is not a common phenomenon. In theory, will result in over four orders of magnitude greater overall galvanic
the galvanic current of AZ91D in the galvanic system ‘‘steel– corrosion damage if the other influences are not considerable,
AZ91D–steel” cannot be simply expressed as the sum of the gal- which may explain the identified corrosion damage in a fuel sys-
vanic current densities of AZ91D in two separate galvanic systems tem and an Al engine cylinder head when E85 is used, whilst no
‘‘steel–AZ91D” and ‘‘AZ91D–steel”. As discussed in Section 3.2, for significant corrosion damage is observed if gasoline is used.
a ‘‘a–b–c” joint system, material (a) can even be influenced by mate- With regard to the maximum galvanic damage at the joint of two
rial (c) and the distribution of IF on it is also determined by the different metals, Eq. (185) predicts (referring to Appendix H.2.1):
parameters of the indirectly connected remote metal (c) in the sys-  c 
tem. However, this ‘‘remote” effect by an indirectly connected re-
E  Eacorr
IaFa  pffiffiffiffiffiffihcorr
pffiffiffiffiffiffi pffiffiffiffiffiffii ð66Þ
mote material is not considered in the ‘‘a–b” and ‘‘b–c” systems. qap qcp þ qap
Therefore, when the galvanic current densities of two simple cou-
ples (‘‘a–b” and ‘‘b–c”) are added together without considering the i.e., the maximum galvanic corrosion rate is mainly determined by
‘‘remote” effect, the ‘‘remote effect” will not be involved in the the polarization resistivity of the coupled metals and the difference
sum of the these two current densities. A lack of the ‘‘remote” effect of their corrosion potentials. Particularly, the polarization resistivity
in the sum of the galvanic current densities should be responsible of a corroding anode is much more critical. This means that in the
for the inadequacy of the superposition principle in the ‘‘a–b–c” in-take valves of an engine block head, improving the polarization
system. resistance of the Al alloy engine block head can more effectively re-
This case study shows that computer modeling or numerical duce the maximum galvanic corrosion damage than increasing the
analysis without analytical verification in some cases can be mis- resistance of the steel seat in theory.
leading, and the analytical approach developed in this study can
help understanding computer simulated results. 5.4. A passive spacer between a galvanic couple

5.3. Corrosion by bio-fuel E85 It is a common practice in the automotive industry to use a pas-
sive spacer to isolate a cathode from an anodic part in order to
Bio-fuel is an important alternative to gasoline and has been reduce the galvanic corrosion. The isolator can be a passive metal,
used in the automotive industry. Currently, the commercial bio- e.g. Al alloy, but more effectively an insulator, such as rubber,
fuel, ‘‘E85” (85 vol% ethanol + 15 vol% gasoline) produced by sev- nylon, or teflon. It is generally accepted in the industry that the
eral oil companies, are available in petrol stations in North America spacer should be wider than 5 mm in order to effectively eliminate
and Europe. However, some corrosion problems associated with the galvanic corrosion attack to Mg alloy parts joined with other
the E85 have been found in fuel systems and aluminum engine cyl- metals. This practice came from Hawke’s publication [27], in which
inder heads. It is unclear what corrosion mechanisms are involved it was reported that the galvanic corrosion indicated by the weight
in the damage. Nevertheless, galvanic corrosion cannot be ex- loss of diecast AZ91D plates was linearly dependent on the insulat-
cluded, as mixed metals are used in these systems. If a galvanic ing spacer thickness (width), and based upon this the linear
corrosion mechanism is assumed, then some corrosion phenomena dependence the galvanic corrosion was predicted to be eliminated
472 G.-L. Song / Corrosion Science 52 (2010) 455–480

if the spacer was thicker (wider) than 4.8 mm. Song et al. [29]
questioned this prediction and experimentally demonstrated that
the galvanic current of AZ91D in contact with steel under the ASTM
B-117 standard salt spray condition was not linearly decreasing
with the spacing width of an insulator and the galvanic current
was found to be still as high as 22 lA/cm2 after the insulating
spacer was up to 9 cm wide. The non-linearly decreasing galvanic
current with spacer width was understood by using a schematic
diagram illustration [29]. Now with the developed galvanic corro-
sion theory in this study, the effect of an insulating spacer on a
galvanic current density can be further analytically examined.
In a salt spray chamber, there should be a thin salt electrolyte
film on specimen surfaces. The studied spacer was an insulator
[29], the polarization resistivity should be infinite. Let AZ91D be
material (a), steel be material (c) and the insulator spacer be mate-
rial (b), then we have qbp ! 1 and Wb ? 0. This is a typical galvanic
ac
system with a small W material between two different metals. Eq. Fig. 19. Dependence of 1=ig on an insulating spacer width. The data points
extracted from the literature (Fig. 8) [29].
(222) (referring to Appendix H.4) can be used to describe the over-
all galvanic current:
immediately. Otherwise, severe local corrosion attack may occur.
ab bc ac Eccorr  Eacorr The localized corrosion attack along a scratch area can be ascribed
ig  ig  ig  pffiffiffiffia pffiffiffiffic
ð67Þ
pffiffiffiffiffi pffiffiffiffiffi q q to the galvanic effect between the exposed substrate along the
qs b qs þ tanhðWp a Þ tanhðWp c Þ
scratch and the surrounding organic coating.
ab If we take the direction perpendicular to the scratch on the
where ig is non-linearly dependent on the spacer width b. This the-
coating surface as the x-coordinate, and assign the scratched sec-
oretically denies the linear dependence of galvanic corrosion rate on
tion as material (b) and the coating on both sides of the scratch
a spacer width [27]. In other words, the existing practice of using
as materials (a) and (c), then a scratched coating can be treated
washers wider than 5 mm to mitigate the galvanic corrosion of
as a one-dimensional system with three pieces in connection. If
Mg alloy parts in the automotive industry may not be sufficient un-
the scratch is made in the middle of a coated specimen, then
der the ASTM standard salt spray environment.
a = c, and the system can be further simplified.
To further support Song et al’s theoretical non-linear argument
Normally, an organic coating itself is nearly an insulator before
[29], the reliability of the dependence of galvanic current on spacer
degradation. It has no electrochemical activity. This means that in
width measured by Song et al. [29] is examined as follows.
a coated area, the polarization resistivity is infinite. When the spec-
Eq. (222) (referring to Appendix H.4) can be rewritten into:
imen is polarized, the resistivity of the coating will be the polariza-
 pffiffiffiffiffi " pffiffiffiffiffiaffi pffiffiffiffifficffi #
1 qs qs qp qp tion resistivity of the coated specimen, which could be over
ac  bþ þ ð68Þ 1014 X cm2. A scratch made by a sharp knife can be as narrow as
ig Eccorr  Eacorr Eccorr  Eacorr tanhðW a Þ tanhðW c Þ
50 lm. If the scratched coating specimen is exposed in a salt spray
which is a straight line and the slope is associated with the electro- environment, the polarization resistivity of the exposed substrate
lyte resistivity and the corrosion potentials of the galvanic system. metal, such as a magnesium alloy, could be only 6.7 X cm2 (see
Fig. 19 presents such a dependence of the galvanic current mea- Fig. 18). At the same time, it has been measured that the solution
sured by Song et al. [29] on the insulator width. It does display a resistance of the salt is about 12.5 X cm (800,000 lS/cm) [25].
nearly straight line. It has been measured [29] that the corrosion Therefore, if the width of the coated area is 1 cm, we can work out
potentials of steel and AZ91D are around 650 mV  Ag/AgCl and the distribution characteristic numbers Wa  Wc  1/(1014/12.5)1/
2
1530 mV  Ag/AgCl, respectively. The slope of the straight line  3.5  107 and Wb  50  104/(6.7/12.5)1/2 = 6.8  103 for
in Fig. 19 is measured to be 0.153 mA1. Therefore, the electrolyte the coated and scratched regions, respectively. Both Wc (or Wa)
resistivity can be calculated to be 134.64 X cm from the slope and and Wb are very small, i.e., the system has small Wa, Wb and Wc.
the corrosion potentials. The width of the tested specimen is al- We can use equations in Appendix H.1.2 to analyze the overall gal-
ready known to be 1.5 cm [29]. If the thickness of the sprayed vanic effect of the coating on the scratch and the maximum galvanic
5 wt% NaCl solution over the specimen is assumed to be around corrosion of the scratched area:
700 lm, then according to Eq. (77), the solution resistivity can be !
bc Eccorr  Eacorr Rbs Ebcorr  Eccorr
estimated, which can be converted into conductivity 82,525 lS/ ig  c  ð69Þ
cm. It has been experimentally measured that this solution has con- Rp þ Rap þ Rbs Rcp Rcp þ Rap þ Rbs
 
ductivity of 79,500 lS/cm [25]. Considering the change in composi- Rap þ Rbs Ebc þ Rcp Eba
tion of the solution due to dissolution of Mg into it and the error in IbFb     ð70Þ
estimating the thickness of the solution on the Mg-steel couple sur- qbp Rcp þ Rap þ Rbs
face under the salt spray condition, the estimated value from Eq.
(68) is very close to the experimental measured one. This suggests It should be noted that material (a) is the same as material (c), so
that the effect of the insulating spacer on galvanic current reported they can be further simplified into:
by Song et al. [29] is reasonably reliable. It appears that the theoret- !
bc Rbs Ebcorr  Eccorr
ical model is useful in verifying experimental results. ig   ð71Þ
Rcp 2Rcp þ Rbs
5.5. A scratch in an organic coating Ebcorr  Eccorr
IbFb   ð72Þ
qbp
Organic coating is one of the most popular corrosion protection
measures in practice. It is common knowledge that if an organic These two equations suggest that the directions of the total galvanic
coating is damaged locally, e.g. scratched, it should be repaired current and the maximum galvanic current density are determined
G.-L. Song / Corrosion Science 52 (2010) 455–480 473

by Ebcorr and Eccorr . A coated area normally


 has a more
 positive corro- where S is the cross-section area of the electrolyte over the metal
bc
sion potential than a scratched area Eccorr > Ebcorr . Hence, ig and IbFb surface; C is the inner perimeter of the tube system and y is the
are positive, flowing from the scratch to the coating. In other words, width of the electrolyte coverage over a flat surface. q0s and q0p have
the scratched area is anodic to the coated area, suffering from a gal- their normal units X cm and X cm2, respectively. They are different
vanic corrosion attack by the coating. from those (qs and qp) in the one-dimensional system.
bc
The ig equation also indicates that the galvanic current from
the scratched area is only dependent on the solution resistance Appendix B
over the scratched region and the polarization resistance over
the coated area. A larger coated area will have a smaller Rcp accord- Based on Fig. 2, there is:
ing to Eq. (74) (referring to Appendix A), and consequently lead to a IF ¼ W=qp ð80Þ
large overall galvanic current over the scratched area. The IbFb equa-
tion suggests that the maximum galvanic current density is mainly where
determined by polarization resistivity of the scratched region in W ¼ ðE  Ecorr Þ ð81Þ
addition to the potential difference between the scratched region where E is a potential of the metal relative to a reference electrode.
and the coated area. Other parameters have no significant influ- Ecorr is the corrosion potential of the metal, which is also relative to
ence on this maximum current density. the same reference electrode. As W is relative to the potential of the
metal, not relative to the potential of the electrolyte or a reference
6. Summary electrode in the electrolyte, its value is negative with respect to the
potential relative to a reference electrode (E  Ecorr). Therefore,
A one-dimensional galvanic system can represent many practi- W ¼ Ecorr  E ð82Þ
cal galvanic corrosion cases. Theoretical equations developed in
Meanwhile, If can be expressed (see Fig. 2) as:
this study for galvanic potentials, current densities and overall gal-
vanic corrosion damage as well as the maximum galvanic damage 1 dW
If ¼  ð83Þ
intensity of typical galvanic systems can be used to analyze many qs dx
real galvanic corrosion problems. The theoretical analytical predic-
At x, the current density flowing into a give elemental electrolyte
tion for galvanic corrosion by this analytical approach provides an
should be equal to the current density flowing out of it (see Fig. 2):
insight into a galvanic corrosion system. With developed theoreti-
cal equations, the influence of system parameters on galvanic cor- IF dx ¼ ðIf jxþdx  If jx Þ ¼ dIf ð84Þ
rosion can be clearly defined.
With Eqs. (83) and (84) can be rewritten:
  
Acknowledgment dw 
dx 
 dw
dx  2
1 xþdx x 1 d w
IF ¼  ¼ ð85Þ
The author thank Dr. Blair Carlson for his help in editing this qs dx qs dx2
document.
According to Eqs. (80) and (85), we have
2
Appendix A d w qs
¼ w ð86Þ
dx
2 qp
The cross-section area of the fine tube and thin electrolyte film
over a flat surface can be written as It has a common solution:
x  x
S ¼ pr 2 ðfor a tubeÞ ð73Þ w ¼ A exp þ B exp  ð87Þ
S ¼ dY ðfor a flat surfaceÞ ð74Þ L L

The solution resistivity qs and polarization resistivity qp in this one- Constant A and B need to be determined with boundary conditions.
dimensional system are: According to Eq. (82), the potential and currents of the metal
relative to a reference electrode can be obtained.
Rs
qs ¼ ð75Þ
m Appendix C
qp ¼ mRp ð76Þ

where Rs is the overall or total solution resistance over the whole According to Eq. (83), there is

region m;Rp is the polarization resistance over the whole region dw
¼0 ð88Þ
m. Solution resistivity qs and polarization resistivity qp are defined dx x¼c
as the solution resistance and the polarization resistance of the me-
tal per unit length, respectively. Both qs and qp are assumed to be This acts as one of the boundary conditions for Eq. (86). The other
constants, independent of the position x in the system. They have boundary condition is the applied potential Wc0 at the other end
units X cm1 and X cm, respectively. Their relationships with the x = 0:
conventional solution resistivity q0s and conventional polarization wc jx¼0 ¼ wc0 ð89Þ
resistivity q0p in a three-dimensional system are:
With these boundary conditions, constants A and B can be easily
q0s determined and the potential and current distributions obtained:
qs ¼ ð77Þ
S    
q0p wc0 wc0 exp  Lcc wc exp  Lcc
qp ¼ ðfor a tubeÞ ð78Þ A¼ 2c ¼ c  c ¼ 0 c ð90Þ
c 1 þ exp Lc exp Lc þ exp  Lc 2 cosh Lc
or      
wc0 exp 2cc wc exp Lcc wc exp Lcc
q0p B¼ L2c ¼  c 0  c ¼ 0   ð91Þ
qp ¼ ðfor a flat surfaceÞ ð79Þ 1 þ exp Lc exp Lc þ exp  Lc 2 cosh c
y L2
474 G.-L. Song / Corrosion Science 52 (2010) 455–480

Therefore, Wc and corresponding Ec ; IcF and Icf equations can be ob- Appendix E
tained according to the general potential and current equations de-
a c
duced earlier. The galvanic current ig or ig between these two pieces of metals
can also be regarded as the polarization currents of metal (a) and
Appendix D metal (c), respectively. If these two metals are jointed at x = 0, they
a c
can be treated as two ‘‘dead end” metals polarized by ig and ig ,
When a ‘‘dead end” is on the right, Icf0 represents the current respectively. According to Eqs. (11) and (95), these are:
flowing into the electrolyte through the left end of the system from 
a waa a
outside, which can also be regarded as an overall polarization cur- ig ¼ ip ¼ Iaf0 ¼  pffiffiffiffiffiffiffiffiffiffi
ffi tanh ð102Þ
qs qpa La
rent ip applied from outside, i.e.: 
  c wc0 c
 wc tanh cc ig ¼ ip ¼ Icf0 ¼ pffiffiffiffiffiffiffiffiffiffi
ffi tanh c ð103Þ
ip ¼ Icf0 ¼ Icf x¼0 ¼ 0pffiffiffiffiffiffiffiffiffifficffiL ð92Þ qs qp c L
qs qp
Substitute them into Eq. (25), then
In theory, the sum of IF over the surface should be equal to the  
polarization current ip from outside. When ip is flowing into the ac wc0 c waa a
ig ¼ pffiffiffiffiffiffiffiffiffiffi
ffi tanh c ¼  pffiffiffiffiffiffiffiffiffiffi
ffi tanh a ð104Þ
electrolyte from left to right, it is considered to be positive. In this qs qp
c L qs qp a L
case, If has to cross the electrolyte/metal interface and flow into
which can be simplified into:
the metal, which is negative. Hence, we have:
pffiffiffiffiffiaffi  
c waa qp tanh Lcc
ip ¼ iF ð93Þ c ¼  pffiffiffiffiffi
ffic a ð105Þ
w0 qp tanh La
Eq. (93) can be confirmed by integrating the Faradic current density
IcF over the entire metal surface. This equation suggests that the potential in the electrolyte over
Eq. (92) can also be rewritten as: metals (a) and (c) are always in opposing direction, i.e., they must
"    # pffiffiffiffiffiffiffiffiffiffiffi be negative to each other.
qffiffiffiffiffiffiffiffiffiffiffi exp cc þ exp  cc ip qs qcp
c According to Eq. (82), we have:
wc0 ¼ iF qs qp c L  L
  ¼   ð94Þ
exp Lcc  exp Lcc tanh Lcc
Ea ¼ Eacorr  Wa ð106Þ
This is a linear equation between applied potential W and current c
0 Ec ¼ Eccorr  Wc ð107Þ
ip.
where Eacorr
and Eccorr
are the corrosion potentials of metal (a) and
If the ‘‘dead end” is on the left, then the applied potential at the
metal (c), respectively. Therefore, Eq. (27) can be rewritten as:
boundary and the responding current Iaf also have a linear relation-
ship as described by Eqs. (92) and (94): Waa  Wc0 ¼ Eacorr  Eccorr ð108Þ
a
waa tanh La  Combination of Eqs. (105) and (108) yields:
ip ¼ iF ¼
a
pffiffiffiffiffiffiffiffiffiffiaffi ¼ Iaf x¼a ¼ Iafa ð95Þ
qs qp waa waa  wc0 Eacorr  Eccorr
pffiffiffiffiffiffiffiffiffiffiffi c 1 ¼ ¼
qffiffiffiffiffiffiffiffiffiffiffi ip qs qap w0 wc0 wc0
a a
wa ¼ iF ¼ qs qp ¼ 
a   ð96Þ pffiffiffiffiffiaffi   pffiffiffiffiffiffi  
tanh Laa qp tanh Lcc þ qcp tanh Laa
¼ pffiffiffiffiffiaffi   ð109Þ
In an ‘‘open ends” system, at both boundaries, the current flowing
qp tanh Lac
into and out of the section can be expressed as: which allows us to work out Waa and Wc0 , respectively:
 
 wb0 cosh b
 wbb  pffiffiffiffiffiffi  
b  Lb Eccorr  Eacorr qap tanh Lcc
i0 ¼ Ibf  ¼ qffiffiffiffiffiffiffiffiffiffiffi   ð97Þ waa ¼  pffiffiffiffifficffi   pffiffiffiffiffiffi   ð110Þ
x¼0
qs qbp sinh b qp tanh Laa þ qap tanh Lcc
Lb
 

 wb0  wbb cosh Lbb  a pffiffiffiffiffiffi  
b
ib ¼ Ibf  ¼ qffiffiffiffiffiffiffiffiffiffiffi   ð98Þ Ecorr  Eccorr qcp tanh Laa
x¼b wc0 ¼  pffiffiffiffiffiaffi  c  pffiffiffiffiffiffi   ð111Þ
qs qbp sinh Lbb qp tanh Lc þ qcp tanh Laa
b  
The total Faradic reaction current iF should be equal to the current The signs of Wc0 and Waa depend on Eacorr  Eccorr . For example, if
density flowing out of the electrolyte minus the current density a c
Ecorr < Ecorr , then c a
W0 > 0 and Wa < 0, and consequently
flowing into the electrolyte over the section at both ends: IaF > 0 and IcF < 0, respectively, according to Eqs. (8) and (15), sug-
b b b
gesting that metal (a) is an anode and metal (c) a cathode.
iF ¼ ib  i0 ð99Þ With the obtained Waa and Wc0 , the potentials and current densi-
Boundary potentials can also be determined if the boundary cur- ties can be expressed according to corresponding Eqs. (6)–(9), (11),
rents are known: (13)–(16), (18), (20)–(23), (97) and (98) developed earlier. It should
qffiffiffiffiffiffiffiffiffiffiffi  8 9 be noted that if the origin point of the x-coordinate is set at the left
b < =
qs qbp sinh b
L
b b
ib  i0
b b
ib þ i0 end of metal (a), then a x-coordinate shift (+a) should be consid-
wb0 ¼  þ   ð100Þ ered in writing equations for metal (c) that has its left end sitting
2 :1  cosh b 1 þ cosh Lbb ;
Lb at x = a. This means ‘‘x” should be replaced with ‘‘x  a” in all the
 8 equations for system (b). Therefore, there are
qffiffiffiffiffiffiffiffiffiffiffi 9
qs qbp sinh b < b b b b =  c pffiffiffiffiffiffi   "  #
b
L ib  i0 ib þ i0 Ecorr  Eacorr qap tanh Lcc cosh Lxa
wbb ¼     ð101Þ wa ¼  hpffiffiffiffiffiffi   pffiffiffiffiffiffi   i a ð112Þ
2 :1  cosh b b ;
Lb
1 þ cosh Lb qcp tanh Laa þ qap tanh Lcc cosh La
G.-L. Song / Corrosion Science 52 (2010) 455–480 475

 "  #
pffiffiffiffiffiffi a pffiffiffiffiffiaffi
c
Eacorr  Eccorr qcp tanh La cosh xac
c
qp
w ¼  hpffiffiffiffiffiffi   pffiffiffiffiffiffi  i
L
  ð113Þ M ¼ qffiffiffiffiffiffi   a ð126Þ
qa tanh cc þ qc tanh aa cosh Lcc q sinh
b b
tanh
p L p L p Lb La
pffiffiffiffifficffi
qp
With them, Ea ; Ec ; IaF ; IcF ; Iaf ; Icf ; IaF0 and IcF0 can all be obtained. N ¼ qffiffiffiffiffiffi   c ð127Þ
b
q sinh
b
p Lb
tanh Lc

b
Appendix F G ¼ 1 þ M cosh ð128Þ
Lb

From the established equations earlier, we have galvanic cur- b
a c H ¼ 1 þ N cosh ð129Þ
rent ig and ig in relation with metals (a) and (c) according to Eqs. Lb
(102) and (103). In the meantime, according to Eqs. (97) and
b b
(98), we have expressions for currents i0 and ib flowing into and With the determined Wb0 and Wbb , Waa and Wc0 will be known accord-
out of the electrolyte over metal (b) at its left and right ends. At ing to Eqs. (120) and (121). Due to the shift of the origin position of
the joints ‘‘a/b” and ‘‘b/c”, the galvanic current flowing from metal x-coordinate, variable x in the above equations needs to be replaced
(a) should be equal to that into metal (b) and that from metal (b) with a shifted x-coordinate.
equal to that into metal (c):
" #  
ab a
ig ¼ ia ¼ i0
b
ð114Þ MEbc þ HEba cosh Lxa
wa ¼  Eba   ð130Þ
bc b c GH  MN cosh Laa
ig ¼ ib ¼ i0 ð115Þ
 
NEba þGEbcbc
þHEba
 MEGHMN exp  Lbb 
Therefore, b GHMN xa
  w ¼   exp
 wb cosh b  wb 2 sinh Lbb Lb
waa a 0 Lb b
  ba bc
 pffiffiffiffiffiffi tanh ¼ qffiffiffiffiffiffi   ð116Þ 
aqp La qp sinh Lb
b b MEbc þHEba
GHMN
exp Lbb NEGHMN
þGE
xa
  þ   exp  b ð131Þ
 wb  wb cosh b 2 sinh Lbb L
wc0 c 0 b Lb
pffiffiffiffifficffi tanh c ¼ qffiffiffiffiffiffi   ð117Þ
qp L qp sinh Lb
b b    
NEba þGEbc
GHMN
 Ebc cosh xabc
Lc
wc ¼ c ð132Þ
They can be rearranged: cosh Lc

pffiffiffiffiffiaffih  i ab
Correspondingly, the values and distributions of ig (Eq. (97)), ig
bc
qp wbb  wb0 cosh b
Lb
waa ¼ qffiffiffiffiffiffi   a ð118Þ (Eq. (98)), Iaf (Eq. (9)), Icf (Eq. (9)), Ibf (Eq. (22)), IaF (Eq. (8)), IcF (Eq.
b
qbp sinh Lb
tanh La (8)), IbF (Eq. (23)), Ea (Eq. (7)), Ec (Eq. (7)) and Eb (Eq. (21)) can all
be obtained according to the equations developed earlier.

pffiffiffiffifficffih  i
qp wb0  wbb cosh b
Lb
Appendix G
wc0 ¼ qffiffiffiffiffiffi   c ð119Þ
b
q sinh
b
p Lb
tanh Lc Let
pffiffiffiffiffiaffi
We also know at the left and right ends of metal (b), the potentials
qp
M ¼ qffiffiffiffiffiffi   a ð133Þ
have the following relationships following the thought of Eq. (108): q sinh
b b
tanh
p Lb La
pffiffiffiffifficffi
qp
wb0  waa ¼ Ebcorr  Eacorr ¼ Eba ð120Þ N ¼ qffiffiffiffiffiffi   c ð134Þ
b
qbp sinh Lb
tanh Lc
wc0  wbb ¼ Eccorr  Ebcorr cb
¼ E ¼ E bc
ð121Þ

then Eqs. (118) and (119) become:


where Eacorr ; Ebcorr and Eccorr are known or measurable. Therefore, 

there are four parameters Waa ; Wb0 ; Wbb and Wc0 which can be deter-
b
waa ¼ M wbb  wb0 cosh b ð135Þ
mined using four available relationships (118)–(121) (refer to L


Appendix G): b
wc0 ¼ N wb0  wbb cosh b ð136Þ
L
ba bc
NE þ GE Substitute
wbb ¼ ð122Þ
GH  MN
Waa ¼ Wb0  Eba ð137Þ
bc ba
ME þ HE c b bc
wb0 ¼ ð123Þ W ¼W E
0 b ð138Þ
GH  MN
which come from Eqs. (120) and (121) into Eqs. (135) and (136),
MEbc þ HEba then they can be written into:
waa ¼  Eba ð124Þ
GH  MN 

b
ba bc wb0 1 þ M cosh b ¼ Mwbb þ Eba ð139Þ
NE þ GE L
wc0 ¼  Ebc ð125Þ 

GH  MN b
wbb 1 þ N cosh b ¼ Nwb0 þ Ebc ð140Þ
where L
476 G.-L. Song / Corrosion Science 52 (2010) 455–480

Let  where
b
G ¼ 1 þ M cosh ð141Þ qap
L
b
Rap ¼ ð160Þ
b a
c
H ¼ 1 þ N cosh ð142Þ qp
Lb Rcp ¼ ð161Þ
c
then
according to Eq. (76). Rap and Rcp are the total polarization resistance
M Eba of metals (a) and (c), respectively.
wb0 ¼ wbb þ ð143Þ
G G
MN b NEba H.1.2. A joint of three pieces of materials
b
wb H ¼ w þ þ Ebc ð144Þ
G b G If a system containing three pieces of metals joined in series is
thus considered, then simplified equations for Waa ; Wb0 ; Wbb and Wc0 can
be developed by using the approximations in Appendix I. If the ori-
NEba þ GEbc gin point or the x-coordinate is set at the left end of metal (a), then
wbb ¼ ð145Þ
GH  MN the x-coordinate shift for metals (b) and (c), must be taken into ac-
count and we have:
Similarly,
   
N Ebc Rap Eacorr  Eccorr þ Rbs Ebcorr  Eacorr
wbb ¼ wb0 þ ð146Þ Ea  Eacorr  waa ¼ Eacorr  ð162Þ
H H Rcp þ Rap þ Rbs
MN b MEbc
b
w0 G ¼ w þ þ Eba ð147Þ  
H 0 H
MEbc þ HEba Rcp þ Rbs Eba þ Rap Ebc
wb0 ¼ ð148Þ b
E  Ebcorr  wb ¼ Ebcorr 
GH  MN Rcp þ Rap þ Rbs
Thus,  
Rb Ea  Ec xa
 s c corr a corr ; a6x6aþb ð163Þ
MEbc þ HEba Rp þ Rp þ Rbs b
waa ¼ ¼ Eba ð149Þ
GH  MN
NEba þ GEbc    
wc0 ¼  Ebc ð150Þ Rcp Eccorr  Eacorr þ Rbs Ebcorr  Eccorr
GH  MN Ec  Eccorr  wc0 ¼ Eacorr  ð164Þ
Rcp þ Rap þ Rbs

   
Appendix H Rap Eccorr  Eacorr þ Rbs Ebcorr  Eacorr
waa
IaF   a ¼   ð165Þ
qp qap Rcp þ Rap þ Rbs
H.1. Small W systems
 
When W ? 0, i.e., Wa ? 0 and Wc ? 0, the approximations as Rcp þ Rbs Eba þ Rap Ebc
wb
described in Appendix I can be employed to simplify the galvanic IbF   b ¼  
qp qbp Rcp þ Rap þ Rbs
potential and current distribution equations.
 
H.1.1. A joint of two dissimilar metals
qbs Eacorr  Eccorr
   ðx  aÞ; a 6 x 6 a þ b ð166Þ
In the case of two different metals joined together, by utilizing qbp Rcp þ Rap þ Rbs
the approximations in Appendix I, we have:
     
Eacorr  Eccorr Rap Rcp Eccorr  Eacorr þ Rbs Ebcorr  Eccorr
a
E  Eacorr  ð151Þ wc0
Rcp þ Rap IcF   c ¼    ð167Þ
  qp qcp Rcp þ Rap þ Rbs
Eccorr Eacorr Rcp
Ec  Ecorr  ð152Þ
Rap þ Rcp    
 a  Rap Eacorr  Eccorr þ Rbs Ebcorr  Eacorr
E  Ecorr c
waa x
IaF   corr  ð153Þ IaF   a ¼    x;
a Rcp þ Rap qp qap Rcp þ Rap þ Rbs
 c 
E  Eacorr 06x6a ð168Þ
IcF   corr  ð154Þ
c Rap þ Rcp  
 a  Rcp þ Rbs Eba þ Rap Ebc
E  Eccorr x Eacorr  Eccorr
IaF   corr  ; 06x6a ð155Þ Ibf  c    x;
a Rcp þ Rap Rp þ Rap þ Rbs qbp Rcp þ Rap þ Rbs
 c 
E  Ea ðx  a  cÞ a6x6aþb ð169Þ
Icf  corr  corr  ; a6x6aþc ð156Þ
c Rap þ Rcp  
 
ac Ec  Eacorr r cp Eccorr  Eacorr þ Rbs Ebcorr  Eccorr
ig  corra ð157Þ Icf     ðx  a  b  cÞ;
Rp þ Rcp qcp Rcp þ Ras þ Rbs
 a 
a Ecorr  Eccorr
IFa     ð158Þ aþb6x6aþbþc ð170Þ
a Rc þ Rap
 c p  !
E  Eacorr  Ea  Ec Rbs Ebcorr  Eacorr
IcF0   corr  ð159Þ ig ¼ Iaf xa   ccorr a corrb 
ab
ð171Þ
c Rap þ Rcp Rp þ Rp þ Rs Rap Rcp þ Rap þ Rbs
G.-L. Song / Corrosion Science 52 (2010) 455–480 477

!  
 Ec  Ea Rbs Ebcorr  Eccorr Ec  Eacorr
ig ¼ Icf x¼aþb   ccorr a corrb 
bc
ð172Þ IaFa  pffiffiffiffiffiffihcorr
pffiffiffiffiffiffi pffiffiffiffiffiffii ð185Þ
Rp þ Rp þ Rs Rcp Rcp þ Rap þ Rbs qap qcp þ qap

     a 
 r ap Eacorr  Eccorr þ Rbs Ebcorr  Eacorr E  Eccorr
IcF0  pffiffiffiffiffiffihcorr ffiffiffiffiffiffii ð186Þ
IaFa ¼ IaF x¼a    ð173Þ p ffiffiffiffiffi
qcp qap þ qcp
ffi p
qap Rcp þ Ras þ Rbs
 
 Rcp þ Rbs Eba þ Rap Ebc H.2.2. Three pieces of materials

IbF0 ¼ IbF     ð174Þ In this case, similarly we can also have simplified expressions
x¼a
qbp Rcp þ Rap þ Rbp for Waa ; Wb0 ; Wbb and Wc0 , and then set the origin point of x-coordi-
nate at the left end  of metal (a)
 to obtain:
 pffiffiffiffiffiffi  
 Rap þ Rbs Ebc þ Rap Eba 2 qap Ebcorr  Eacorr a x
 a
E  Eacorr þ qffiffiffiffiffiffi pffiffiffiffiffiffi exp  a cosh a ;
IbFb ¼ IbF     ð175Þ L L
x¼aþb
qbp Rcp þ Rap þ Rbp qp þ qp
b a

06x6a ð187Þ
  
qffiffiffiffiffiffi 
 rcp Eccorr  Eacorr
þ Rbs
Ebcorr  Eccorr 
IcF0 ¼ IcF x¼aþb    ð176Þ b b
qbp Ebcorr  Eccorr xac
qcp Rcp þ Ras þ Rbs E  Ecorr  qffiffiffiffiffiffi pffiffiffiffiffiffi exp
qbp þ qcp Lb
qffiffiffiffiffiffi 
where qbp Ebcorr  Eacorr 
xa
þ qffiffiffiffiffiffi pffiffiffiffiffiffi exp  b ; a 6 x 6 a þ b ð188Þ
Rbs ¼ bqs ð177Þ qbp þ qap L

according to Eq. (75). Rbs is the total electrolyte resistance along


pffiffiffiffiffiffi 

2 qcp Ebcorr  Eccorr c
metal (b). Ec  Eccorr þ qffiffiffiffiffiffi pffiffiffiffiffiffi exp  c
qbp þ qcp L

H.2. Large W systems xabc
cosh ; aþb6x6aþbþc ð189Þ
Lc
a c
When W ? 1, i.e., W ? 1 and W ? 1, the approximations as
 
described in Appendix J and be employed to simplify the galvanic  
2 Ebcorr  Eacorr a x
potential and current distribution equations. IaF  pffiffiffiffiffiffiqffiffiffiffiffiffi pffiffiffiffiffiffi exp  a cosh a ;
qp qp þ qp
a b a L L
H.2.1. A joint of two dissimilar metals
06x6a ð190Þ
By utilizing the approximations in Appendix J and setting the
2
origin point at the left end of system (a), we have: 
 pffiffiffiffiffiffi
 1 6 Eb  Eccorr xab
2  Eccorrq Eacorr x a IbF   qffiffiffiffiffiffi 4qcorr
ffiffiffiffiffiffi pffiffiffiffiffiffi exp
Ea  Eacorr þ pffiffiffiffifficffi pffiffiffiffiffiaffi expðW a Þ cosh a ;
p
qbp qbp þ qcp Lb
qp þ qp L 3

06x6a ð178Þ Eb  Eacorr xa 7
þ qcorr
ffiffiffiffiffiffi pffiffiffiffiffiffi exp  b 5; a 6 x 6 a þ b ð191Þ
 pffiffiffiffiffiffi  qbp þ qcp L
2 Eacorr  Eccorr qcp xac
Ec  Eccorr þ pffiffiffiffiffiaffi pffiffiffiffifficffi expðW c Þ cosh ;
qp þ qp Lc  
2 Ebcorr  Eccorr  
a6x6aþc ð179Þ c xabc
IcF  pffiffiffiffiffiffiqffiffiffiffiffiffi pffiffiffiffiffiffi exp  c cosh ;
  qcp qbp þ qcp L Lc

2 Eccorr  Eacorr x
IaF  pffiffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffii expðW a Þ cosh a ;
h 06x6a ð180Þ aþb6x6aþbþc ð192Þ
qp qp þ qp
a c a L
 
  2 Ebcorr  Eacorr  
 a x
2 Eacorr  Eccorr xac Iaf   pffiffiffiffiffiqffiffiffiffiffiffi pffiffiffiffiffiffi exp  a sin a ;
IcF  pffiffiffiffiffiffihpffiffiffiffiffiffi pffiffiffiffiffiffii expðW c
Þ cosh ; qs qbp þ qap L L
qcp qap þ qcp Lc
06x6a ð193Þ
a6x6aþc ð181Þ
2
   
2 Eccorr  Eacorr x 1 6 Ebcorr  Eccorr xab
If  pffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffii expðW a Þ sinh a ;
a
h 06x6a ð182Þ IbF   pffiffiffiffiffi 4qffiffiffiffiffiffi pffiffiffiffiffiffi exp
L qs qbp þ qcp Lb
qs qp þ qpc a
3

   Eb  Eacorr xa 7
2 Eacorr  Eccorr xac  qcorr
ffiffiffiffiffiffi pffiffiffiffiffiffi exp  b 5; a 6 x 6 a þ b ð194Þ
Icf  pffiffiffiffiffihpffiffiffiffiffiffi pffiffiffiffiffiffii expðW c
Þ sinh ; qbp þ qap L
qs qap þ qcp Lc
 
a6x6aþc ð183Þ 2 Ebcorr  Eccorr  
c xabc
Icf   pffiffiffiffiffiqffiffiffiffiffiffi pffiffiffiffiffiffi exp  c sin ;
  qs qbp þ qcp L La
ac Eccorr  Eacorr
ig  pffiffiffiffiffi pffiffiffiffiffiffi pffiffiffiffiffiffii
h ð184Þ
qs qap þ qcp aþb 6 x 6 aþbþc ð195Þ
478 G.-L. Song / Corrosion Science 52 (2010) 455–480

ab
 Eb  Eacorr H.4. A small W material between two other materials
ig ¼ Iaf x¼a  pffiffiffiffifficorr
qffiffiffiffiffiffi pffiffiffiffiffiffi ð196Þ
qs qbp þ qap
This is an extreme case when Wb ? 0, but Wa – 0 and Wc – 0.
Utilizing the approximations in Appendix I, we can work out the
bc
 Eb  Eccorr galvanic potentials and current densities as well as the overall gal-
ig ¼ Icf x¼aþb   pffiffiffiffifficorr
qffiffiffiffiffiffi pffiffiffiffiffiffi ð197Þ
vanic currents and maximum current densities:
qs qbp þ qcp
La
 Eb  Eacorr Eac 
b sinh ðW a Þ x
IaFa ¼ IaF x¼a  pffiffiffiffiffifficorr
qffiffiffiffiffiffi pffiffiffiffiffiffi ð198Þ wa  cosh ; 06x6a ð211Þ
La
a c
qap qbp þ qap 1 þ b tanhL þ b tanhL
ðW a Þ ðW c Þ

 Eb  Eacorr La
Ebc
ba
þ bE þ tanhL W c Eba
c
 Eacorr  Eccorr
IbF0 ¼ IbF    qffiffiffiffiffifficorr
qffiffiffiffiffiffi pffiffiffiffiffiffi ð199Þ b
w 
tanh ðW a Þ ð Þ
þ ðx  aÞ;
x¼a
qbp qbp þ qcp La Lc
b þ tanh W a þ tanh W c
a c
b þ tanhL W a þ tanhL W c
ð Þ ð Þ ð Þ ð Þ
a6x6aþb ð212Þ
 Eb  Eccorr
IbFb ¼ IaF x¼aþb   qffiffiffiffiffifficorr
qffiffiffiffiffiffi pffiffiffiffiffiffi ð200Þ
Lc
qbp qbp þ qcp Eca 
c b sinh ðW c Þ xabc
w  a c cosh c ;
1 þ b tanhL þ b tanhL L
ðW c Þ ðW c Þ
 Eb  Eccorr
IcF0 ¼ IcF x¼aþb  pffiffiffiffiffifficorr
qffiffiffiffiffiffi pffiffiffiffiffiffi ð201Þ aþb6x6aþbþc ð213Þ
qcp qbp þ qcp
Ea Ec x
The simplified equations predict reasonable results. IaF   corr corr pffiffiffiffi pffiffiffiffic
cosh a ;
pffiffiffiffiffiaffi a pffiffiffiffiffi qap qp L
b qp sinh ðW Þ qs þ tanh W a þ tanh W c
ð Þ ð Þ
H.3. A joint of a large W material and a small W material
06x6a ð214Þ
If Wa ? 1 and Wc ? 0, after mathematical approximations, we
2 La bc ba a 3
have: a E þbE þ tanhL W c Eba a c
1 tanh ðW Þ ð Þ E E
pffiffiffiffiffiffiffiffiffiffiffi   IbF  b 4 þ corr corr
ðxaÞ5;
2 qs qap Eccorr  Eacorr
a c a c
a x qp bþ tanhL W a þ tanhL W c bþ tanhL W a tanhL W c
E  Eacorr þ c pffiffiffiffiffiffiffiffiffiffiaffi expðW a Þ cosh  a ; ð Þ ð Þ ð Þ ð Þ
Rp þ qs qp L
a6x6aþb ð215Þ
06x6a ð202Þ
 c  Ec Ea
Rcp
Ecorr  Eacorr IcF   corr corrpffiffiffiffi pffiffiffiffic

pffiffiffiffiffiffi pffiffiffiffiffi qa q
Ec  Eccorr  c pffiffiffiffiffiffiffiffiffiffiaffi ð203Þ b qcp sin ðW c Þ qs þ tanh Wp a þ tanh Wp c
Rp þ q s q p ð Þ ð Þ
xabc
pffiffiffiffiffi   cosh ; aþb6x6aþbþc ð216Þ
2 q Ec  Ea x Lc
IaF ffiffiffiffiffiaffi expðW a Þ cosh a ;
  a psffiffiffiffiffi corr c pcorr 06x6a ð204Þ
qp qs þ Rp qp L 
Ea  Eccorr x
  Iaf   corr pffiffiffiffia pffiffiffiffic
sinh a ;
pffiffiffiffiffi pffiffiffiffiffi qp qp L
wc0 Rcp Eccorr  Eacorr Ec  Eacorr a
qs sinhðW Þ b qs þ tanhðW a Þ þ tanhðW c Þ
IcF ¼    p ffiffiffiffiffiffiffiffiffiffi
ffi  ¼  corr pffiffiffiffiffiffiffiffiffiffi
ffi ð205Þ
qcp qcp Rcp þ qs qap c Rcp þ qs qap 06x6a ð217Þ
  
2 Eccorr  Eacorr x Eccorr  Eacorr
Iaf   pffiffiffiffiffiffiffiffiffiffi
ffia expðW a
Þ sinh ; 06x6a ð206Þ Ibf   pffiffiffiffia pffiffifficffi
qs qp þ Rcp La pffiffiffiffiffi pffiffiffiffiffi q q
qs b qs þ tanhðWs a Þ þ tanhðWs c Þ
wc0 Ec  Eacorr L a ba
Ebc þ bE þ tanhðW
L
c E
ba c

Icf   ðx  a  cÞ ¼  corrpffiffiffiffiffiffiffiffiffiffi
ffi ðx  a  cÞ; a6x tanhðW a Þ
 qffiffiffiffiffiffih
Þ
i ðx  aÞ; a6x6aþb ð218Þ
q c
p Rcp þ qs qap c La Lc
qbp b þ tanhðW a þ
Þ tanhðW c Þ
6aþc ð207Þ
 
Eccorr  Eacorr sinh xabc
La
Icf   pffiffiffiffi pffiffiffiffic
;
pffiffiffiffiffi pffiffiffiffiffi qa q
ab
 Ec  Eacorr Eccorr  Eacorr qs sinhðW c Þ b qs þ tanhðWp a Þ þ tanhðWp c Þ
ig ¼ Iaf x¼a  pcorr
ffiffiffiffiffiffiffiffiffiffiaffi qcp ¼ pffiffiffiffiffiffiffiffiffiffi
ffi ð208Þ
qs qp þ c qs qap þ Rcp aþb 6 x 6 aþbþc ð219Þ
pffiffiffiffiffi c 
 qs Ecorr  Eacorr  Eacorr  Eccorr
ig ¼ Iaf x¼a  
ab
IaF x¼a   a p ffiffiffiffiffi c pffiffiffiffiffi
ffi ð209Þ pffiffiffiffia pffiffiffiffic
ð220Þ
qp qs þ Rp qap pffiffiffiffiffi pffiffiffiffiffi q q
qs b qs þ tanhðWp a Þ þ tanhðWp c Þ
 Ec  Eacorr
IcF x¼a    corr pffiffiffiffiffiffiffiffiffiffi
ffi ð210Þ bc
 Eccorr  Eacorr
c Rcp þ qs qap ig ¼ Icf0 þ Icf x¼aþb  pffiffiffiffia pffiffiffiffic
ð221Þ
pffiffiffiffiffi pffiffiffiffiffi q q
qs b qs þ tanhðWp a Þ þ tanhðWp c Þ
The distributions of potentials and currents are highly asymmetric.
G.-L. Song / Corrosion Science 52 (2010) 455–480 479

ab bc ac Eccorr  Eacorr H.5. A large W material between two small W materials


ig  ig  ig  pffiffiffiffia pffiffiffiffic
ð222Þ
pffiffiffiffiffi pffiffiffiffiffi q q
qs b qs þ tanhðWp a Þ þ tanhðWp c Þ This is an extreme case when Wa ? 0 and Wc ? 0. Utilizing the
approximations in Appendix I, we can work out the galvanic poten-
 Ea  Eccorr tials and current densities as well as the overall galvanic currents
IaFa ¼ Iaf x¼a   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi corr pffiffiffiffia pffiffiffiffic
ð223Þ and maximum current densities:
pffiffiffiffiffi q q
b qap tanhðW a Þ b qs þ tanhðWp a Þ þ tanhðWp c Þ
Rap Eba
2 3 Ea  Eacorr  qffiffiffiffiffiffiffiffiffiffiffi ð242Þ


L bc a
Lba ba
1 tanhðW a Þ E þ bE þ tanhðW c Þ E 5
c
qs qbp þ Rap
IbF0 ¼ Ibf   b4 La Lc
ð224Þ
x¼a qp b þ tanhðW a þ
Þ tanhðW c Þ qffiffiffiffiffiffiffiffiffiffiffi
qs qbp Ebc 
xab
2 a
bc
3 Eb  Ebcorr  qffiffiffiffiffiffiffiffiffiffiffi exp
 Lc

L bc
1 4tanhðW a Þ E þ bE þ tanhðW c Þ E 5
ba
qs qbp þ Rcp Lb
IbFb ¼ Ibf   b La Lc
ð225Þ qffiffiffiffiffiffiffiffiffiffiffi
x¼aþb qp b þ tanhðW a þ
tanhðW c Þ 
Þ qs qbp Eba xa
 qffiffiffiffiffiffiffiffiffiffiffi exp  b ; a 6 x 6 a þ b ð243Þ
 Ec  Eacorr qs qbp þ Rap L
IcF0 ¼ IcF x¼aþb   qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi corr pffiffiffiffia pffiffiffiffic
ð226Þ
pffiffiffiffiffi q q
b qcp tanhðW c Þ qs þ tanhðWp a Þ þ tanhðWp c Þ Rcp Ebc
Ec  Eccorr  qffiffiffiffiffiffiffiffiffiffiffi ð244Þ
In the case that the metals (a) and (b) have extremely large W, e.g. qs qbp þ Rcp
Wa ? 1 and Wc ? 1, the above potential and current equation can
be further simplified: Eba
a  Iaf   qffiffiffiffiffiffiffiffiffiffiffi  ð245Þ
2L ðEacorr  Eccorr Þ x a qs qbp þ Rap
wa  a expðW a Þ cosh a ; 0 6 x 6 a
c ð227Þ
bþL þL L
ba  a 2 qffiffiffiffiffiffiffiffiffiffiffi qffiffiffiffiffiffiffiffiffiffiffi
La Ebc þ bE þ Lc Eba E  Ec b bc 
wb  a c þ corr a corrc ðx  aÞ; 1 6 qs qp E xab qs qbp Eba
bþL þL bþL þL IbF ¼  b 4qffiffiffiffiffiffiffiffiffiffiffi exp þ qffiffiffiffiffiffiffiffiffiffi

a6x6aþb ð228Þ qp qs qbp þ Rcp Lb qs qbp þ Rap
 

2Lc ðEccorr  Eacorr Þ a xabc xa


wc  expðW Þ cosh ; exp  b ; a6x6aþb ð246Þ
b þ La þ Lc Lc
L
aþb6x6aþbþc ð229Þ

2La ðEa  Ec Þ x Ebc
IaF   a corr a corr c expðW a Þ cosh a ; 0 6 x 6 a ð230Þ Icf   qffiffiffiffiffiffiffiffiffiffiffi  ð247Þ
qp ðb þ L þ L Þ L c qs qbp þ Rcp
ba
La Ebc þ bE þ Lc Eba Ea  Ec
IbF   a c  b corr a corr c ðx  aÞ; x Eba
qp ðb þ L þ L Þ
b qp ðb þ L þ L Þ Iaf   qffiffiffiffiffiffiffiffiffiffiffi ð248Þ
a q qb þ Ra
a6x6aþb ð231Þ s p p

c 2Lc ðEc  Ea Þ c xabc 2 qffiffiffiffiffiffiffiffiffiffiffi
IF   c corr a corr expðW Þ cosh ;
qp ðb þ L þ Lc Þ Lc qs qbp Ebc 
b 1 6 xab
aþb6x6aþbþc ð232Þ IF ¼  qffiffiffiffiffiffiffiffiffiffiffi 4qffiffiffiffiffiffiffiffiffiffiffi exp
 qs qbp qs qbp þ Rcp Lb
a 2ðEacorr  Eccorr Þ a x
IF   expðW Þ sinh ; a6x6aþb ð233Þ qffiffiffiffiffiffiffiffiffiffiffi 3
qs ðb þ La þ Lc Þ La qs qbp Eba 
ba xa 7
Eacorr  Eccorr La Ebc þ bE þ Lc Eba  qffiffiffiffiffiffiffiffiffiffiffi exp  b 5; a 6 x 6 a þ b ð249Þ
IbF   a c  ðx  aÞ; qs qbp þ Rap L
qs ðb þ L þ L Þ qbp ðb þ La þ Lc Þ
a6x6aþb ð234Þ
 
2ðEccorr  Eacorr Þ xabc xabc Ebc
IcF   a c expðW Þ sinh
c
c ; Icf   qffiffiffiffiffiffiffiffiffiffiffi ð250Þ
qs ðb þ L þ L Þ L c q q b þ Rc s p p
aþb6x6aþbþc ð235Þ
 Eacorr  Eccorr 
ig ¼ Iaf x¼a  
ab
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi ð236Þ ab  Eb  Eacorr
bqs þ qs qap þ qs qcp ig ¼ Ibf0 ¼ Ibf   qcorr
ffiffiffiffiffiffiffiffiffiffiffi ð251Þ
 Eccorr  Eacorr
x¼a
qs qbp þ Rap
ig ¼ Icf x¼aþb 
bc
pffiffiffiffiffiffiffiffiffiffiffi pffiffiffiffiffiffiffiffiffiffiffi ð237Þ
bqs þ qs qap þ qs qcp
 Eb  Eccorr
 Ea  Eccorr bc 
IaFa ¼ IaF x¼a   pffiffiffiffiffiffiffiffiffiffiacorr
ffi pffiffiffiffiffiffiffiffiffiffiffi ð238Þ ig ¼ Ibfb ¼ Ibf    qcorr
ffiffiffiffiffiffiffiffiffiffiffi ð252Þ
b qs qp þ qap þ qap qcp x¼aþb
qs qbp þ Rcp
 p ffiffiffiffiffi
ffi p ffiffiffiffiffi pffiffiffiffiffiffi
 qap Ebc þ b qs Eba þ qcp Eba
IbF0 ¼ IbF     pffiffiffiffi pffiffiffiffi  ð239Þ
x¼a
qbp b qs þ qap þ qcp
pffiffiffiffi  Eb  Eacorr
IaFa ¼ IaF x¼a   qcorr
ffiffiffiffiffiffiffiffiffiffiffi  ð253Þ
 pffiffiffiffiffiaffi bc pffiffiffiffiffi pffiffiffiffiffiffi a qs qbp þ Rap
 qp E þ b qs Ebc þ qcp Eba
IbFb ¼ IbF    pffiffiffiffi pffiffiffiffi pffiffiffiffi  ð240Þ
x¼aþb
qbp b qs þ qap þ qcp pffiffiffiffiffi b 

 qs Ecorr  Eacorr
 Ec  Eacorr IbF0 IbF 
IcF0 ¼ Ic    pffiffiffiffiffiffiffiffiffifficcorr
ffi pffiffiffiffiffiffiffiffiffiffiffi ð241Þ
¼ ¼ pffiffiffiffiffi qffiffiffiffiffiffi ð254Þ
F x¼aþb
b qq
s p þ p þ q c a c
p p qq
x¼a
qbp qs þ qbp Rap
480 G.-L. Song / Corrosion Science 52 (2010) 455–480

pffiffiffiffiffi b  
c 1

c

 qs Ecorr  Eccorr cosh ! exp
IbFb ¼ IbF  ¼ pffiffiffiffiffi qffiffiffiffiffiffi ð255Þ Lc 2 Lc
x¼aþb
qbp qs þ qbp Rcp    
a 1 a b 1 b
cosh  a ! exp a cosh  b ! exp b
L 2 L L 2 L
 
 Eb  Eccorr c 1 c
IcF0 ¼ IcF x¼0   qcorr
ffiffiffiffiffiffiffiffiffiffiffi  ð256Þ cosh  c ! exp c
c qs qbp þ Rcp L 2 L
  
a b c
tanh a ! 1 tanh b ! 1 tanh c ! 1
L L L
  
Appendix I a b c
tanh  a ! 1 tanh  b ! 1 tanh  c ! 1:
L L L
When W ? 0, there are:
m x
!0 !0 References
L   L x
m m x
exp !1þ exp !1þ [1] E. Blasco-Tamarit, A. Igual-Munoz, J.G. Anton, et al., Corrosion Science 51
L
 m L L x L x (2009) 1095.
m [2] K. Fushimi, A. Naganuma, K. Azumi, et al., Corrosion Science 50 (2008) 903.
exp  !1 exp  !1
mL L L L [3] R.M. Souto, Y. Gonzalez-Garcia, A.C. Bastos, et al., Corrosion Science 49 (2007)
m x x 4568.
sinh ! sinh ! [4] J.-M. Lee, Electrochimica Acta 51 (2006) 3256.
 L m L m L  L
x x [5] A.P. Yadava, A.P. Yadav, H. Katayama, K. Noda, H. Masuda, A. Nishikata, T.
sinh  ! sinh  ! ð257Þ Tsuru, Electrochimica Acta 52 (2007) 3121.
m L L x L L [6] C. Andrade, P. Garces, I. Martinez, Corrosion Science 50 (2008) 2959.
cosh !1 cosh !1 [7] W.T. Tsai, J.R. Chen, Corrosion Science 49 (2007) 3659.
 L m L 
x
[8] F. Andreatta, H. Terryn, J.H.W. de Wit, Corrosion Science 45 (2003) 1733.
[9] X.G. Zhang, E.M. Valeriote, Corrosion Science 34 (1993) 1957.
cosh  !1 cosh  !1 [10] X.G. Zhang, Corrosion 56 (2000) 139.
m L x L x
m [11] C. Wagner, Journal of the Electrochemical Society 98 (1951) 116.
tanh ! tanh ! [12] J.T. Waber, Journal of the Electrochemical Society 101 (1954) 271.
 L m L m L  L
x x
[13] J.T. Waber, Journal of the Electrochemical Society 102 (1955) 344.
tanh  ! tanh  ! [14] J.T. Waber, Journal of the Electrochemical Society 102 (1954) 420.
L L L L [15] J.T. Waber, B. Fagan, Journal of the Electrochemical Society 103 (1956) 64.
[16] J.A. Simmons, S.R. Coriell, F. Ogburn, Journal of the Electrochemical Society 114
where m can be a, b, or c, the length of system (a), (b) or (c). L can be (1967) 782.
La, Lb, and Lc. W can be Wa, Wb or Wc. [17] E. Kennard, J.T. Waber, Journal of the Electrochemical Society 117 (1970) 880.
[18] L. Gal-Or, Y. Raz, J. Yalhalom, Journal of the Electrochemical Society 120 (1973)
598.
Appendix J [19] E. McCafferty, Journal of the Electrochemical Society 124 (1977) 1869.
[20] J.O. Dukovic, C.W. Tobias, J.T. Waber, Journal of the Electrochemical Society
101 (1954) 271.
When W ? 1, there are: [21] Y.G. Kim, Y.-C. Kim, Y.-T. Kho, Boundary Element Technology XIV (3) (2001) 88.
[22] K.S. Yeung, Boundary Element Technology XIV (3) (2001) 273.
a b c
!1 !1 !1 [23] M. Verbrugge, Corrosion Science 48 (2006) 3489.
La Lb Lc [24] J. Jia, G. Song, A. Atrens, Corrosion Science 48 (2006) 2133.
   [25] J. Jia, G. Song, A. Atrens, Materials and Corrosion 11 (2004) 55.
a b c
exp a ! 1 exp b ! 1 exp c ! 1 [26] A. Tahara, T. Kodama, Corrosion Science 42 (2000) 655.
L L L [27] H. Hack, Materials Performance 18 (1989) 72.
   [28] J.R. Scully, H.P. Hack, Prediction of tube–tubesheet galvanic corrosion using
a b c
exp  a ! 0 exp  b ! 0 exp  c ! 0 finite element and wagner number analyses, in: H.P. Hack (Ed.), Galvanic
L L L Corrosion, ASTM, 1988, pp. 136–157.
    [29] G. Song, B. Johannesson, S. Hapugoda, D. StJohn, Corrosion Science 46 (2004)
a 1 a b 1 b
sinh a ! exp a sinh b ! exp b 955.
L 2 L L 2 L [30] G.-L. Song, Recent progressin corrosion and protection of Mg alloys, Advanced
 
c 1 c Engineering Materials 7 (7) (2005) 563.
sinh c ! exp c [31] G.-L. Song, Corrosion and Protection of Mg alloys, Chemical Industry Press,
L 2 L 2006. pp. 38–40 (in Chinese).
   
a 1 a b 1 b [32] G.L. Makar, J. Kruger, Corrosion of Mg, International Materials Reviews 38
sinh  a !  exp a sinh  b !  exp b (1993) 138.
L 2 L L 2 L [33] J.P. De Souza, O.R. Mattos, L. Sathler, H. Takenouti, Corrosion Science 27 (1987)
 
c 1 c 1351.
sinh  c !  exp c [34] M. Scholz, J. Ellermeier, Materialwissenschaft und Werkstofftechnik 37 (2006)
L 2 L 842.
   
a 1 a b 1 b
cosh a ! exp a cosh b ! exp b
L 2 L L 2 L

You might also like