Journal of Nuclear Materials 392 (2009) 396–404

Contents lists available at ScienceDirect

Journal of Nuclear Materials

journal homepage: www.elsevier.com/locate/jnucmat

Determination of the long-term intergranular corrosion rate

of stainless steel in concentrated nitric acid
V. Bague, S. Chachoua, Q.T. Tran, P. Fauvet *
CEA, DEN, DPC, SCCME, Laboratoire d’Etude de la Corrosion Non Aqueuse, F-91191 Gif-sur-Yvette, France

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

Article history: Stainless steels with low carbon content and free from any precipitation undergo intergranular attack in
Received 17 December 2008 hot nitric acid. The corrosion rate measured by weight loss requires prolonged immersion testing to reach
Accepted 19 December 2008 the apparent steady state corrosion, which coincides with the onset of grain dropping. A more appropri-
ate method for predicting the long-term penetration rate is described in this study. A close observation
and a statistical analysis of the attack grooves were firstly undertaken using immersion testing. The major
PACS: findings are an outstanding morphology of the grooves with flat planes and preserved angle even after
81.05.Bx
the onset of grain dropping, as well as a constant rate of the penetration into the surface. The formation
81.70.q
82.45.Bb
of the grooves could then be represented by a geometrical model put forward by Beaunier and co-work-
ers. Consequently, the method proposed for predicting the penetration rate consists in measuring the
depth and the angle of the grooves obtained in short time immersion testing. Multiplying the penetration
rate calculated from the previous data by the ratio between the penetration depth and the length of the
grain boundary path does give an accurate long-term penetration rate. The method has been shown to
apply successfully to AISI 304L stainless steel in several nitric solutions.
Ó 2009 Elsevier B.V. All rights reserved.

1. Introduction grain dropping. A method for predicting the long-term corrosion

rate has been suggested by these authors. It consists in evaluating
Intergranular corrosion is observed on austenitic stainless steels the weight loss of specimens that undergo pre-corrosion treatment
in hot nitric acid even in the absence of chromium carbide precip- inducing grain dropping. The method allows deconvolution of the
itates. This type of degradation could be related to the structure corrosion rate with respect to each exposed face by the application
and energy of the grain boundaries, as well as to the segregation of an empirically-determined factor.
of some alloying elements in these areas [1–6]. Two states appear The present study details another method to predict the long-
to exist during the propagation of the corrosion: the corrosion rate term corrosion rate of stainless steel in hot nitric acid from
calculated from the weight loss increases with immersion time un- short-term tests. In order to take into account the localized nature
til a steady state is reached [7–9] which coincides with the onset of of the attack as well as to overcome the difficulties to determine
grain dropping (Fig. 1). In most cases, at the beginning, the corro- the real area of the metal undergoing intergranular corrosion, the
sion rate is too slow to reach the steady state with short time corrosion rate will be straightforwardly calculated from the depths
experiments performed in laboratory. Therefore, predictions from of the attack path.
results obtained with short-term tests could lead to an important
underestimation of the long-term corrosion rate. 2. Experimental details
However, Dunned, Whillock and co. [10–13] emphasized re-
cently that the weight loss is hitherto normalized with respect to Specimens were machined from a 20 mm thick plate from a sin-
the nominal surface area of the test specimen, instead of to the ac- gle heat of AISI 304L stainless steel with low carbon content and
tual area of the metal undergoing corrosion. The latter increases free of chromium carbide precipitates, the composition of which
with time and reaches a constant value after the onset of stable is given in Table 1. Grains are equiaxial. The mean grain size D
grain dropping. Consequently, according to these authors, the determined according to the French standard NF A04-102 is about
apparent increase in the corrosion rate obtained gravimetrically 60 lm. Each specimen is a 30  20  1.5 mm3 parallelepiped. Its
is an artefact. Thus, measurement of the intergranular corrosion surface is wet ground to 2400 grit using SiC paper and is finally
rate by weight loss is correct only if the metal is undergoing stable polished to a mirror finish using 1 lm diamond powder.
Experiments are immersion tests in nitric media. The test
* Corresponding author. Tel.: +33 1 69 08 20 24; fax: +33 1 69 08 15 86. solutions used in this study were sufficiently oxidizing to reach
E-mail address: [email protected] (P. Fauvet). rapidly the steady state corrosion rate, by adding Ce4+ oxidizing

0022-3115/$ - see front matter Ó 2009 Elsevier B.V. All rights reserved.
doi:10.1016/j.jnucmat.2008.12.100
 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404 397

3. Results
Weight loss Δm

Figs. 3–5 show the surfaces of the specimens after various test-
ing times. For every examined specimen, the morphology of the at-
tack has the same features over the whole surface. It suggests that
propagation of intergranular corrosion is isotropic. The main fea-
tures of the degradation are as follows:

1. All grains boundaries are attacked as well as twinning planes

(Fig. 3).
0 50 100 2. Opened grooves are formed at boundaries. They have outstand-
Time (d) ing geometry with two stable planes as sides (Fig. 4). These
planes form an angle a which bisector plane is likely the grain
Fig. 1. Weight loss vs. time observed on stainless steel undergoing intergranular boundary surface. Confirmation of these features is apparent
corrosion in hot nitric acid. on the cross-sections of the grooves which are always triangular
and well defined (Fig. 5).
3. Other parameters which characterize the groove dimensions
Table 1 are indicated in Fig. 6. The cases of a dissymmetrical and a sym-
Chemical Composition (wt%) of 304L stainless steel.
metrical groove are, respectively, presented on the left and on
Fe C Ni Cr Mn Cu Si P S B Co the right of the figure. The parameters H and L (respectively,
Bal 0.017 10.16 18.34 1.69 0.116 0.57 0.024 0.001 0.0002 0.018 H0 and L0 ) are the depth and the width of the groove if the initial
surface is taken as the reference (respectively, if the final mean
surface is taken as the reference).
Table 2 H (respectively, H0 ) which is the projection of H (respectively,
Nitric media. H0 ) on the plane perpendicular to the initial surface, corre-
sponds to the penetration depth. In the case of a symmetrical
Solution A 8 M HNO3 + 1 g/L Ce4+
Solution B 8 M HNO3 + 5 g/L Ce4+ groove,
H ¼ H ;
H0 ¼ H0 :
ions (standard potential E°25 °C = 1610 mV/SHE for the Ce4+/Ce3+
couple). The two test solutions are listed in Table 2. Fig. 2 depicts
the apparatus used for the immersion tests. Twenty two specimens
were suspended in 2000 mL of test solution so that the ratio of the
surface area of exposed metal and the liquid volume (i.e. the S/V
ratio) is 0.148 cm1. The solution is continuously renewed with a
42 mL h1 flow rate whereas its temperature is maintained at
100 °C.
One specimen is periodically removed and replaced by a new
coupon in order to keep constant the specified S/V ratio. The re-
moved specimen is weighed to determine its weight loss and is
examined by optical and scanning electron microscopes. An
exhaustive statistical analysis of the parameters of the grooves
formed at grain boundaries, i.e., depth, width and angle, is per-
formed from 100 grooves ‘nearly symmetrical’ (see Section 3) ran-
domly picked up on the specimen cross-section.
Fig. 3. Surface of 304L SS specimen after 24 h in solution A at 100 °C.

Vapor trap

Thermocouple

Solution
in

Pump

Specimens
Tank
containing test
Solution Cell solution
out

Heater Balance
Fig. 4. Observed groove at grain boundary on a 304L SS specimen in solution A at
Fig. 2. Schematic immersion test apparatus. 100 °C.
398 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404

Fig. 5. Cross-section micrograph of a 304L specimen after 7 days in solution B at 100 °C.

Grain boundary
plane
Initial surface L L
H* H H* = H
Final surface L’ L’

H’* H’= H’*

H’
α α
Dissymmetrical groove Symmetrical groove

Fig. 6. Schematic groove dimensions.

Symmetrical groove should yield a maximal penetration depth as 4. Discussion

compared with dissymmetrical groove, as for the same value of H,
the former has a higher H than the latter. Consequently, only 4.1. Formation of groove at grain boundaries according to Beaunier’s
nearly ‘symmetrical’ grooves with respect to the plane perpendic- model
ular to the initial surface (angle difference less than 5°), will be se-
lected for a close statistical analysis of the groove dimensions. The features of the grooves observed in this study are similar to
The groove depths H0 and the groove angles a follow a Gaussian those observed by Beaunier and co. on 304L stainless steel in 1 M
distribution characterized by their mean value and their standard
deviation during the immersion time (Figs. 7 and 8). It suggests
25
that the advancing front of the attack is regular. This feature is
actually observed on the specimen’s cross-sections (Fig. 5).
4. In addition, during the immersion test, the groove depth H0 is 20
never higher than the grain size of the material. It means that
the attack doesnot affect deeply the material, in contrast with
Occurrence

15
the morphology of end-grain corrosion [13]. The groove angle
is being constant and is roughly 25° for the two test solutions
10
(Figs. 9 and 10).

For a distance shorter than the mean grain size, the groove 5

depth H experiences a linear increase with time immersion

(Fig. 11). In addition, a linear change with time of the groove depth 0
H is also observed for the distance longer than the mean grain size. 9.7 13.5 17.2 20.9 24.6 28.4 32.1 35.8 39.5 43.2 45.6

However, the slope of the corresponding plot is smaller. Fig. 12 Angle α (°)
shows that the weight loss increases also linearly with the immer-
Fig. 8. Distribution of groove angle a after 120 h in solution A.
sion time.

45
25
40

20 35

30
Angle (°)
Occurrence

15 25

20
10
15

10
5
5

0
0
0.8 1.8 2.8 3.7 4.7 5.7 6.7 7.6 8.6 9.6 10.5 0 500 1000 1500 2000 2500 3000 3500 4000 4500

Depth H’ (µm) Immersion time (hours)

Fig. 7. Distribution of groove depth H0 after 120 h in solution A. Fig. 9. Groove angle vs. time in solution A.
 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404 399

45 sulphuric acid under potentiostatic polarisation [14–16]. These

40 authors have proposed an electrochemical model for intergranular
corrosion under transpassive conditions. It is as if there are two
35
constant dissolution vectors in equilibrium: the penetration rate of
30 grain boundary Vg and the dissolution rate of the groove sides
Angle (°)

25 and of the specimen surface VS (Fig. 13). Stable grooves are indeed
formed because Vg is higher than Vs. Relation (1) could be then
20
established:
15
H 1
10 ¼ ; ð1Þ
e sin a2
5
e is the thickness loss due to the uniform corrosion and, as seen la-
0
0 50 100 150 200 250 300 ter in the chapter, to the grain dropping process.
Immersion time (hours) As H is defined by

Fig. 10. Groove angle vs. time in solution B.

H ¼ H0 þ e; ð2Þ
Relation (1) gives:
H0
H¼ : ð3Þ
1  sin a2
200

180 Fig. 14 shows the evolution of the groove depth H calculated from
relation (3) using experimental data H0 and a for the distance short-
Penetration depth H (µm)

160
er than the mean grain size. Since these values match those ob-
140
tained in this study, the propagation of the intergranular
120
corrosion seems to fit Beaunier’s dissolution model.
100 This type of attack along the grain boundaries must lead to the
80 grain dropping during the propagation of the corrosion. The weight
60

40
Original
surface
20

VS d
e
0
0 2000 4000 6000 8000 10000 12000 L’
Final
Immersion time (hours)
Surface
Solution B
α/2
Solution A Vg H’
Linear regression from Hexperimental
data obtained with short time tests
Linear regression from all Hexperimental data H

Fig. 11. Observed mean penetration depth H* vs. time.

Fig. 13. Schematic groove formation according to Beaunier’s model.

120
60

100
50
Penetration depth H (µm)
Weight loss (mg.cm-2)

80 40

60 30

Δm(1 row)
40 20

20 10

0 0
0 2000 4000 6000 8000 10000 12000 14000 0 100 200 300 400 500 600 700 800

Immersion time (hours) Immersion time (hours)

Solution B H* predicted in solution B

Solution A H* experimental in solution B
Predicted weight loss using a grain H* predicted in solution A
modelled as a cube H* experimental in solution A
Linear regression from weight loss data
Fig. 14. Observed penetration depth and predicted penetration depth as a function
Fig. 12. Weight loss vs. time. of time.
400 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404

loss could then be predicted in the following way: equiaxial grain 35

could be represented as a regular cube or a truncated octahedron.

30
Other identical units are then packed to fill the space as shown in
Fig. 15. When the advancing front of the intergranular corrosion 25

Occurrence
reaches values between 1.5 D (in the case of a grain shape mod-
elled as a cube) and 1.8 D (in the case of a grain shape modelled 20

as a truncated tetrahedron), one row of grains is broken away from

15
the surface of the specimen. This period of time t varies from 5860
to 6840 h for the solution A (8 M HNO3 + 1 g/L Ce4+) and from 1390 10
to 1680 h for the solution B (8 M HNO3 + 5 g/L Ce4+). The corre-
sponding weight loss is about 47.5 mg/cm2. The specimen loses 5

one entire row of grains every such interval of time. As the esti- 0
mated weight loss is actually consistent with a good accuracy with 2 11 21 30 40 50 60 69 79 88
the experimental weight loss data (Fig. 12), grains should drop row dH/dt *103 (µm.h-1)
by row during the propagation of the intergranular corrosion. dHshort time
Grains are actually observed at the bottom of the corrosion cell Fig. 17. Distribution of dt experimental
in solution A.

(Fig. 16).
 short time
The penetration rate dH is given by the groove depth-
dt A H 1
time plot obtained from short time immersion testing, which re- ¼ ; ð4Þ
e sin a2
quires statistical analyses of groove depths. It could also be pre-
dicted with only one immersion test. Since the intergranular dH 1 de
¼ : ð5Þ
corrosion rate agrees with Beaunier’s dissolution model, we obtain dt sin a2 dt
with our experimental results:
For an angle a between 0 and 60°, sin a2 is not very different from
tan a2. It can then be replaced by tan a2 in Relation (5) with an error
inferior to 15%:
dH 1 de 1 de 1 de
¼ ffi ¼ 0 ; ð6Þ
dt sin a2 dt tan a2 dt L =20 dt
H
H* H
Or
dH H0 de
D ¼ L0 : ð7Þ
dt 2
dt

In addition,
de dH
t þ H0 ðtÞ ¼ t : ð8Þ
dt dt
The combination of (7) and (8) gives:

dH 2½H0 ðtÞ2
¼ : ð9Þ
dt ½2H0 ðtÞ  L0 ðtÞt
 short time
Therefore, the penetration rate dH
dt A
could be determined by
Fig. 15. Difference between the penetration depth H* and the grain boundary path’s relation (10) from one immersion test after a lapse of time t1:
length H.  short time
dH 2½H0 ðt 1 Þ2
¼ : ð10Þ
dt A ½2H ðt1 Þ  L0 ðt 1 Þt1
0

Such Gaussian distributions of H0 (t) and L0 give a Gaussian distribu-

 short time  short time
tion of dHdt experimetal
(Fig. 17). As seen in Table 3, the dH
dt predicted
obtained is consistent with the slope resulting from the groove
depth-time plot.
In conclusion, these results show that the formation of the
grooves at grain boundaries agrees with the Beaunier’s dissolution
model; the propagation of the grooves leads to the periodical ‘row
by row’ grain dropping. Such a type of propagation yields a linear
increase in penetration depth H, which is usually observed by other
authors [10,12,18].

Table 3
 short time
Comparison of dH
dt predicted
(lm/h) determined from groove depth-time plot and from
Beaunier’s model.
dHshort time dHshort time
dt predicted
from groove dt predicted
from one groove depth
depth-time plot according to Beaunier’s model
Solution B 0.12 ± 0.008 0.21 ± 0.06
Solution A 0.036 ± 0.002 0.055 ± 0.015
Fig. 16. Grain dropped from 304L SS specimen.
 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404 401

4.2. Determination of the relationship between the penetration depth 350

H0 and the grain boundary path’s length H0 affected by intergranular

300

Penetration depth H* (µm)

corrosion
250
Since a linear increase in thickness loss indicates a constant cor-
rosion rate, prediction of the intergranular corrosion rate from 200

groove depth data appears quite easy. However, as seen in

150
Fig. 11, the corrosion rate determined from the short time immer-
short time
sion testing dH dt experimetal
differs from the corrosion rate determined
long time 100
from long time immersion testing dH dt experimetal
. It is illustrated by
the fact that the slopes of the corresponding groove depth-time 50
short time
plots are not similar (Fig. 11). The dH dt experimetal
is about twice higher
long time
than the dHdt experimetal
. This result could be straightforwardly associ- 0
0 2000 4000 6000 8000 10000 12000
ated with the difference between the penetration depth H0 and Time (hours)
the length of the grain boundary path H0 . This idea can be illus- H* predicted in solution B using method 1
trated in Fig. 15 using a very simple two dimensional representa-
H* experimental in solution B
tion of the grain shape modelled as a cube.
By analogy to the terminology used in the porous media, a cor- H* predicted in solution A using method 1 + linear regression
relation factor k, labelled as ‘tortuosity’ factor, could be defined as: H* experimental in solution A + linear regression

H0 Fig. 19. Penetration depth H predicted using method 1.

k¼ : ð11Þ
H0
about ten corresponding grain boundary paths H0 were measured
It represents the ratio between the penetration depth H0 (in other
(Fig. 18). Using the mean value of H0 , we obtain:
words the thickness of the metal affected by the intergranular cor-
rosion) and the length of the grain boundary path H0 (in other words H0
k¼ ¼ 0:70  0:01: ð13Þ
the real path followed by the intergranular corrosion). In all cases, H0
this ratio must be less than 1 and is expected to be unique for a Fig. 19 shows the penetration depths calculated by using the
dHshort time
homogeneous material. The depth penetration could then be simply dt
and the factor k. This method provides a better account
estimated from Relation (12): of the experimental
dHshort time data than does the method using only
 short time dt
. However, the difference between the calculated slope
dH and the experimental slope of the penetration depth-time plot is
H¼k t: ð12Þ
dt still quite important, around 67%.
Several methods mainly based on geometrical considerations are
examined in order to determine the factor k. 4.2.2. Determination of the tortuosity factor k – Method 2
The factor k is determined using a representation of grain mod-
4.2.1. Determination of the tortuosity factor k – Method 1 elled as truncated octahedron (Fig. 20). For the mean grain size D,
This method consists in measuring the mean grain boundary the side a of the octahedron is given by:
path observed on a cross-section of the specimen prepared for D
a ¼ pffiffiffiffiffiffi : ð14Þ
metallographic examination. For a distance H0 randomly chosen, 10
The longest grain boundary path H0 , represented by the dotted line
in Fig. 20, corresponds to the penetration depth H0 represented by
the unbroken line. These parameters are given by:
pffiffiffi
2 þ 2 cosð35 16Þ
H0 ¼ pffiffiffiffiffiffi D; ð15Þ
10
5
H0 ¼ pffiffiffiffiffiffi D; ð16Þ
10
pffiffiffi
H0 2 þ 2 cosð35 16Þ
k¼ 0 ¼ : ð17Þ
H 5

Fig. 18. Cross-section of the specimen. Fig. 20. Grain modelled as truncated octahedron.
402 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404

 short time
350
dH
H0 predicted ¼ t: ð18Þ
300
dt B
Penetration depth H (µm)

It gives:
250
H0 experiment
k¼ ¼ 0:34  0:07: ð19Þ
200
H0 predicted
150
The predicted penetration depth is in agreement with the experi-
mental data (Fig. 22).
100

50 4.2.4. Comments
The factor k obtained from these three methods and the
0 dHlong time dHlong time
dt predicted
corresponding are reviewed and compared to dt experimental
0 2000 4000 6000 8000 10000 12000

Immersion time (hours) in Table 4 for the solution A. The second and third method give a
long time
quite accurate prediction for the dH dt
.
H* predicted in solution B using method 1
dHshort time
H* predicted in solution B using method 2 As seen in §4.1, the dt predicted can also be determined using
H* experimental in solution B relation (10) according to the Beaunier dissolution’s model. By
applying the factor k determined using the third method to this va-
H* predicted in solution A using method 1  long time
lue, we obtain dH for the solution A which is also consistent
H* predicted in solution A using method 2 dHshort timedt predicted
with dt experimental data (Table 5 and Fig. 23).
H* experimental in solution A
We must emphasize that the factor k determined from the sec-
Fig. 21. Penetration depth predicted using method 2 – comparison with method 1. ond method doesn’t vary with the mean grain size for materials
with equiaxial grains. It may indicate that for a material with equi-
axial grains, the factor k defined in this study does not depend on
the grain size. This information is not opposed to the ideas widely
350
accepted in the literature, which are put forward that grain size has
300 an important influence on the intergranular corrosion rate. In fact,
Penetration depth H (µm)

for a given volume of metal, a variation of the grain size induces a

250 variation in grain boundary surface so that the higher is the grain
size, the smaller is the intergranular surface. The impurity concen-
200
tration is constant in the material but impurities are shared out be-
150
tween grain boundary surfaces. Consequently, in most cases, the
impurity concentration in the grain boundaries increases with
100 the grain size, so is the intergranular corrosion rate. In our study,
the grain size may be without influence on the factor k but the
50 short time
opposite is expected on dH dt predicted
. Further experiments are planned
0
to examine the effect of the grain size upon the factor k and the
dHshort time
0 2000 4000 6000 8000 10000 12000 dt predicted
, although we are aware that grain size and impurity con-
Immersion time (hours) centration might not be independent parameters. Actually, any
cold working and heat treatment used for increasing the grain size
H* predicted in solution A using method 1
influences the segregation of impurity in the grain boundaries.
H* predicted in solution A using method 2 The results obtained in this study are also in agreement with
H* predicted in solution A using method 3 Dunnett’s statements on the apparent increase of corrosion rate
observed from weight loss data [10]. According to these authors,
H* experimental in solution A

Fig. 22. Penetration depth predicted using method 3 – comparison with methods 1
and 2. Table 4
 long time
Comparison of dH
dt predicted
(lm/h) calculated using the factor k for the solution A.

Method 1 Method 2 Method 3 Experimental

So, the factor k is 0.57. Fig. 21 shows the penetration depth-time
 short time Factor k 0.70 ± 0.01 0.57 0.28 ± 0.14 _
plot predicted from the factor k and the value of dH
dt
. The dif- dHshort time
_ _ _ 0.036 ± 0.002
dt experimental
ference between predicted slope and experimental slope of the plot dHlong time
dt predicted
0.025 ± 0.002 0.021 ± 0.001 0.010 ± 0.003 _
is not greater than 33%.
dHlong time
dt experimental
_ _ _ 0.015 ± 0.001

4.2.3. Determination of the tortuosity factor k – Method 3

The factor k could also be determined experimentally using a
nitric media reference sufficiently oxidizing so that the phase of Table 5
dHlong time
grain dropping is quickly attained, for example within a month. dt predicted
(lm/h) determined using k ratio calculated from method 3 and using
dHshort time
A solution containing 8 M HNO3 and 5 g/L of Ce4+ (solution B) sat- dt predicted
calculated from Beaunier’s model (i.e. the relation (10)).
isfies such a condition for 304L stainless steel. Solution B Solution A
A specimen was immersed in this medium and only removed dHshort time
dt predicted
0.215 ± 0.06 0.055 ± 0.015
after losing a thickness higher than the mean grain size D. The fac- k 0.30 ± 0.15 0.30 ± 0.15
tor k could then be calculated from the value of H0 experimental mea- dHlong time
0.060 ± 0.03 0.014 ± 0.003
dt predicted
sured on the specimen and the value of H0 predicted predicted by the dHlong time
dt experimental
0.063 ± 0.03 0.015 ± 0.001
relation (18):
 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404 403

350

300

Penetration depth H (µm)

250

200

150

100

50

0
0 2000 4000 6000 8000 10000 12000
Immersion time (hours)

H* predicted in solution A using method 3 and

dH short time
determined by the relation
dt predicted
(10)
H* predicted in solution A using method 3 and
dH short time
dt exp erimental
H* experimental in solution A

Fig. 23. Penetration depth predicted in solution A (8 M HNO3 + 1 g/L Ce4+) using the factor k determined from method 3.

the apparent increase in corrosion rate obtained gravimetrically is indicates that the intergranular corrosion rate is constant during
an artefact since the weight loss is hitherto normalized with re- the propagation of intergranular corrosion.
spect to the nominal surface area of the test specimen, instead of Finally, the method 2 could be used for materials with equiaxial
to the actual area of the metal undergoing corrosion. The latter in- crystallization while the method 3 may be used for any material.
creases with time up to reach a constant value when grain drop-
ping steady state is achieved (Fig. 24). This idea is not easily 5. Conclusion
conceivable. But, using a two dimensional representation, a simu-
lation of the advancing front according to Beaunier’s dissolution This study shows that the propagation of intergranular corro-
model seems to justify a preservation of the surface area of the sion of stainless steel such as 304L in hot nitric acid leads to the
advancing front after the stable grain dropping, i.e. after the drop- formation of grooves with outstanding geometry: triangular
ping of the first row of grains. As seen in Fig. 25, the surface areas of grooves with regular faces, preserved angle during the propagation
the interface at t ¼ 32 VDg ; t ¼ 2 VDg and t ¼ 52 VDg are identical. Ohno and of the intergranular corrosion. This morphology is consistent with
co. obtained the same type of interface when they simulate the the grain boundary dissolution model proposed by Beaunier and
weight loss promoted by intergranular corrosion [17]. However coll. It is as if there are two constant dissolution vectors in equilib-
from our point of view, this result is conclusive only when the rium, the penetration rate vector of grain boundary Vg and the dis-
material is subject to intergranular corrosion and not to end-grain solution rate of the groove faces and of the surface of the
corrosion [13]. specimens VS.
Even if the apparent increase of the weight-loss rate is not seen Statistical analyses as well as an examination of the cross-sec-
in the present study, a linear increase of the penetration depth tion of the specimen reveal a regular advancing front. These fea-
tures of the degradation agree with a grain dropping ‘row by
row’ process. The grain dropping process and the linear increase
Preserved surface after in groove depth observed on the distance shorter than the mean
stable grain dropping grain size should result in increasing linearly the penetration
Initial depth, which is actually observed in this study.
Consequently, the penetration depth could be simply determined
surface
from the short-term penetration rate obtained from short time tests
by the application of a factor k, labelled as ‘tortuosity’ factor; the lat-
ter takes into account the ‘real’ grain boundary path. On one hand,
the short-term penetration rate could be calculated either from the
groove depth-time plot or from the Beaunier’s dissolution model.
On the other hand, the factor k could be determined experimentally
or by modelling grain shape as truncated octahedron.
Work is in progress to determine whether grain size has a great
influence on the factor k and on the penetration rate obtained from
Fig. 24. Schematic surface undergoing intergranular corrosion. short-term tests data.
404 V. Bague et al. / Journal of Nuclear Materials 392 (2009) 396–404

t=0 D
t=
Vg
i 0 1 2 3 4
→ →
( 0, i, j ) D 0.8 D 0.6 D 0.4 D 0.2 D

→ →
'
(0 , i, j )

D
3 D D
t= t=2
2 Vg Vg

5 D
t=
2 Vg

Fig. 25. Simulation of the advancing front for stainless steel undergoing intergranular corrosion in hot nitric acid.

Acknowledgements of International Symposium on Plant Aging and Life Predictions of Corrodible

Structures, Sapporo, Japan, 1995, p. 853.
[9] M. Mayuzumi, J. Ohta, T. Arai, Corrosion 54 (1998) 271.
This work has been supported by AREVA NC and CEA authori- [10] B.F. Dunnett, G.O.H. Whillock, Corrosion 59 (2003) 274.
ties. We are also grateful to D. Mas, B. Fieulaine, R. Robin, B. Gwin- [11] M. Takeuchi, G.O.H. Whillock, J. Nucl. Sci. Technol. 41 (2004) 702.
[12] G.O.H. Whillock, B.F. Dunnett, Intergranular corrosion testing of austenitic
ner, F. Balbaud and A. Terlain for their helpful suggestion and
stainless steels in nitric acid solution, in: Proceedings of Eurocorr 2004, Nice,
assistance during this study. France, 2004.
[13] G.O.H. Whillock, B.F. Dunnett, M. Takeuchi, Corrosion 61 (2005) 58.
[14] L. Beaunier, M. Froment, D. Lepoutre, C. Vignaud, Métaux Corros. Industrie 610
References (1976) 211.
[15] L. Beaunier, M. Froment, C. Vignaud, Electrochim. Acta 25 (1980)
[1] J.S. Armijo, Corrosion 24 (1968) 24. 1239.
[2] H. Coriou, A. Desestret, L. Grall, J. Hochman, Mémoires scientifiques Rev. [16] L. Beaunier, Journal de Physique, Colloque C6, supplement to Journal de
Métallurg LXI (3) (1964) 177. Physique 43 (12) (1982) C6-271. Tome.
[3] A. Desestret, I. Epelboin, M. Froment, P. Guiraldenq, Corros. Sci. 18 (1968) 225. [17] A. Ohno, H. Isoo, M. Akashi, Numerical model for intergranular corrosion rate
[4] A. Desestret, J. Ferriol, G. Vallier, Mater. Technol. 65 (9–10) (1977) 621. of stainless steel in nitric acid, in: Proceedings of International Symposium on
[5] J. Stolarz, Journal de Physique IV, Colloque C7, supplement to Journal de Plant Aging and Life Predictions of Corrodible Structures, Sapporo, Japan, 1995,
Physique III 5 (1995) C7-423. p. 869.
[6] M. Mayuzumi, J. Ohta, K. Kako, Corrosion 56 (2000) 70. [18] H. Coriou, L. Grant, G. Kurka, G. Plante, Influence des ions chrome sur la
[7] H. Coriou, J. Hure, G. Plante, Electrochim. Acta 5 (1961) 105. corosion intergranulaire d’aciers inoxydables austénitiques en milieu
[8] M. Mayuzumi, T. Fujita, Effect of oxidizing ions and heat transfer on the nitrique, in: Proceedings of Colloque de Métallurgie, Saclay, France, Vol. IV,
corrosion rates of stainless steels and zirconium in nitric acid, in: Proceedings 1960, p. 75.