Académique Documents
Professionnel Documents
Culture Documents
P. G. Shewmon
To cite this article: P. G. Shewmon (1985) Hydrogen attack of pressure-vessel steels, Materials
Science and Technology, 1:1, 2-11, DOI: 10.1179/mst.1985.1.1.2
Article views: 29
NELSON CURVES (API, 1977) several samples under a fixed pressure but at increasing
temperatures. A complete loss in ductility is usually found
over a temperature increment of 50 K.
600 3. Pressure. Hydrogen attack is not observed in carbon
II steel until the hydrogen partial pressure PH exceeds 1 MPa .
.\
C-)
At higher pressures the incubation time decreases slowly
~500
0::
1\. \ 2.25 Cr -1.0 Mo with increasing pressure, or equivalently the temperature
:;::)
t--
I ', . threshold for HA drops (see Fig. 1).
~LoU 400
a...
::E
LoU
t-- 300
< ------- O~MO
---
Steel
4. Alloying. Alloying with carbide-forming elements can'
significantly increase the steel's resistance to HA, that is,
the steel resists attack to a higher temperature. 3,5 The
.- .- .- _ . _. CARBON Steel elements most commonly added for this purpose are Cr
200
-'- and Mo. Reducing the carbon content is not an effective
10 15 20 MPa
approach since steels with only a few hundredths per cent
o 1 2 3 ksi of carbon are still attacked.2, 17 .
HYDROGEN PARTIAL PRESSURE 5. Microstructure. In carbon steel no changes are
Lines suggested by API as .upper limits of tempera-
obvious under the light microscope during the incubation
ture and pressure for use of various steels5 .
stage. However, work with the scanning electron
microscope (SEM) has shown that early in the incubation
period discrete bubbles develop on grain-boundary
carbides (see Fig. 2). These grow and ultimately link up to
Observations form thin fissures along the grain boundariesl8,19 (see
Fig. 3). These fissures in turn link together, and often vent
Downloaded by [Universite Laval] at 09:13 09 May 2016
properties, and next most severe for the ductility, while the bubbles. At lower temperatures where methane formation
ultimate tensile strength and the yield stress are last and is slower, and/or at later stages where the bubbles link up
least affected.2,15 to form fissures, the existence of the equilibrium pressure in
In unstressed, quenched and tempered (QT) 2'25Cr-1Mo the bubbles is less clear. However, during the long
steel Erwin24 reported no change in the room-temperature incubation time before bubble link-up the bubbles grow
tensile properties after 5 years at 540C and 21 MPa slowly and it seems quite probably that equilibrium is a
hydrogen even though a bubble volume fraction of between good approximation, at temperatures above 275-300C.
001 and 01 had developed. Under these stress-free Several experimental observations point to this.19,28-30
conditions bubbles form and grow, but fissuring does not The methane that generates HA in steel is formed by the
occur. However, Cuiffreda et al.25 showed that 138 MPa reaction
of hydrogen at 455C reduced the creep ductility
MxC+2H2=CH4+xM (1)
significantly in a test that failed in 6 months. Also, Wanagel
et a1.26 found that modest exposure to hydrogen (14 MPa, For Fe3C in carbon steel the activity of the carbide and the
600C, 20 d) developed methane bubbles sufficient to metal can be taken as unity. The equilibrium constant K
reduce the stress-rupture life materially in subsequent tests for this reaction can be written as
in an inert atmosphere at 600C. Both of these creep tests
were done under hydrogen-exposure conditions that would
1m = KIJ . . . . . . . . . . . . . . . (2)
have developed much less bubble volume by HA alone where 1m and IH are the fugacities of methane and
than was developed in Erwin's long-time tests. hydrogen, and Ao and AM are the activities of carbon and
7. Decarburization. The depth of decarburization in the metal (iron) atoms. For Fe3C, 1m is given by the
normal service 450C) should be negligible unless equation
fissuring develops to aid transport to and from the ambient
In/m=-13'3+13700/T+2InPH (3)
gas. In Cr-Mo steels exposed to hydrogen Masaoka
Downloaded by [Universite Laval] at 09:13 09 May 2016
et ai.2 7 reported surface decarburization and 'cracks' along where the fugacity of the hydrogen gas has been taken
grain boundaries near the surface. At 450C decarburiza- equal to its partial pressure, i.e~ hydrogen is assumed to
tion was negligible. Even after 1000 h at 600C the depth of behave ideally.31
decarburization was only 0'3-04 mm. The depth of grain- Odette and Vagarali32 have developed an equation of
boundary cracking (fissuring) at 600C increased with the state for methane in this range of temperature and
depth of decarburization and with PH' but was less than pressure. They show that 1m is adequately described by an
the depth of decarburization. equation of the form
(4)
where
Mechanisms C(T) = 2'375/T + 178 x 10- 3 1m > 104 MPa }
= 1187/T+309 X 10-3 (5)
Lengthening the incubation time is the critical factor in 103 <1m < 104 MPa
preventing HA. The use of sensitive dilatometry and the = 0005 1m < 103 MPa
SEM have established that during essentially all of the
The combination of these equations leads to relations
incubation period many small (submicrometre) methane
between PH2 and PCH4 as shown in Fig. 4. Note that at low
bubbles are growing on the grain boundaries of the stee1.19
hydrogen pressures and higher temperatures the lines on
Thus an analysis of the kinetics of this early period
this log-log plot approach the slope of Pm ~ P~, as would
becomes an analysis of the rates of nucleation and growth
be expected for ideal gas behaviour, while at high hydrogen
of these small bubbles. The pressurized bubbles create
pressures and lower temperatures the non-ideality of the
discontinuities in the stress normal to the grain boundaries
methane leads to a nearly constant methane pressure,
that results in a stress gradient exactly the same as that
independent of PH2 This high-pressure behaviour is a
caused by the stress discontinuities formed by voids on
strong contributor to the nearly pressure-independent
boundaries in samples under an external load. Thus the
temperature limit for HA found in Fig. 1.
atomic processes involved in the nucleation and growth of
methane bubbles in steel closely resemble the slow growth
of grain-boundary voids during the creep of alloys, with the
formation of fissures in HA corresponding to the fracture
of the creep sample. 10 ksi Equilib. PMVS PH / II
/ II
Since the methane pressure provides the driving force for d -- [ar bon Steel 1/1
both nucleation and growth, the discussion of mechanisms ~a.. --- Cr-Mo Steel 1//
w
a
1.5 II I /
begins with the calculation of the maximum (equilibrium) 0::: / I /
:::> 3 ksi
methane pressure that can develop as a function of VI
VI
// / /
hydrogen pressure, temperature, and carbide type (carbon w
activity).
0:::
a..
z
w 1ksi
Ac= O.12~~//
/'/'
//> / / "71
II
~a ,/ / /
EQUiliBRIUM METHANE PRESSURE 0:::
/' / /
.",/' ,/,/ /
It is impossible to draw any valid conclusions about the 0
>- 0.5 /
/ / /
/' ,/
mechanism of bubble growth in HA unless the pressure :I: / /
.", /'
driving the expansion of the bubbles is known. The early ~a 0.3 ksi
/
/
.",
/
/'
/'
/'
/ /
--I ,/ .", ./
kinetic models assumed that the methane pressure equalled 0 0/ 550C ,/ 500 /(./450 500C
the methane fugacity. This led to such high driving 0.5 1.5 2 2.5 3 3.5
pressures that one always concluded that bubble growth LOG (METHANE PRESSURE7MPa)
occurred by matrix creep.
4 Methane pressure in equilibrium with steel as
The bubbles in which the methane forms are initially
function of hydrogen pressure and temperature:
very small, and growth is very slow. Thus, it seems clear solid lines are for carbon steels; broken lines are for
that at higher temperatures, and early in the incubation carbon activity 012 and reflect lower carbon
period, the equilibrium pressure is established in the activity found in commercial QT 2'25-1 Mo steel
0.3
2 Carbon Activity
(Graphite Std. State)
>-
.....
~ E:20.3 MPa H2 >
"0~ 0 E: \6.9 MPa H2
~
U
X
0.2
~.c z
. 0
w dl
0:::
2
U
0.\
Downloaded by [Universite Laval] at 09:13 09 May 2016
19 20 21
T (20+ I"ogt) x 10-3
5 Activity of carbon, and dilatometric strain rates at 550C as function of tempering parameter36
Alloy steels is, much more iron IS In the carbides than assumed by
In steels alloyed with carbide-forming elements, like Geiger and Angeles. These high Fe/Cr ratios stem
chromium, the activity of carbon is lower than in carbon primarily from the low chromium content of the steel.
steels. Also, dilatometric measurements have established However, it is partly caused by the lack of chromium equi-
that the rate of HA of QT 2'25Cr-IMo is at least three libration between the carbides and regions of the alloy far
orders of magnitude slower than that of carbon steels at from the carbides owing to the slow diffusion of chromium.
the same temperature and hydrogen pressure. 1 7.33 Several
observations indicate the effect of alloying and heat GROWTH KINETICS
treatment on the microstructure and rate of HA: Models
The growth of methane bubbles in a steel requires that'
1. The line on the Nelson curve (see Fig. 1) for
several processes proceed simultaneously:
2'25Cr-lMo is about 200 K higher than for carbon steel.
2. Geiger and Angeles31 found, as have several other (i) the solution and diffusion of hydrogen
authors, 3.34 that the resistance to HA in quenched Cr-Mo (ii) the solution and diffusion of carbon
steels increased appreciably with tempering. They also (iii) the formation of methane gas
concluded that the thermodynamic properties of the more (iv) the movement of iron atoms away from the growing
stable M23C6 and M7C3 carbides formed by tempering bubble.
could be approximated by using the free energy of
formation of Cr23C6 and Cr7C3. From this they concluded Figure 7 shows the steps that must occur sequentially, or in
parallel, for the growth of methane bubbles. Anyone of
that the activity of carbon should drop by roughly three
these processes could be rate controlling, but analysis40-44
orders of magnitude as tempering removed the M3C from
the steel.
3. Baker and Nutting35 report that during extended
tempering of QT 2'25Cr-IMo steel at 690C all of the
M3C disappears and is replaced by M23C6 or M7C3.
Parthasarathy and Shewmon36 have recently measured o XT 1 50
6 XT2
the carbon activity of QT 2'25Cr-IMo steel at 550C as a -15 o XT3
function of the amount of tempering. They did this by XT4 CD
Q
o 8M
equilibrating the steel with pure Ni.37 Results are shown in 1: )(
Parallel Process I
( bl
Surface Diffusion
of Fe Atoms
Bubble Growth
Towards Bubble Tip
by Direct Power
t Law or Dislocation
Diffusion of Fe Creep
Atoms Along the
Grain Boundary
(c I
I
I
Accommoda t i on by I
Creep of the I
I
Adja~ent Grains I I I I
II I I
r--- aj-----+-- f3-l--a
Downloaded by [Universite Laval] at 09:13 09 May 2016
t:cr = (fb b/d) B((3/2n)/(b/a)3/0 -lJo po. . . . . . (10) inferences can be drawn about the rate-controlling
processes if the rate data are described with the equation
Note that the force is raised to the power n, which in the
case of iron is '" 7. E=AP~exp(-Qm/RT) (12)
Dyson has noted that voids form on only a fraction of
the grain boundaries in a sample undergoing creep, and Note that Qm reflects only the temperature dependence of
that if this fraction is small enough a cage of grains with the diffusion process limiting the motion of iron atoms,
uncavitated boundaries will surround any cavitated while QH contains both the temperature dependence of the
boundary.48 The cavitated boundary and the surrounding diffusion process and the dependence of methane pressure
cage of void-free grains must expand at the same rate, so on temperature at constant hydrogen pressure. The
the cage may restrain the growth of the cavitated pressure exponent m in equation (12) should be directly
boundary. It is possible that the same cage effect would comparable with the exponents of P in the model rate
occur in restraining the growth of discrete bubble-covered equations, equations (7), (8), or (10).
grain boundaries in samples undergoing HA, especially in Carbon steel Several steels have been studied using
the more creep and HA resistant alloys like the Cr-Mo dilatometry and the data fit with equation (12). Domains of
steels. different rate-controlling processes have been observed. If
Whether or not this type of restraint occurs is an the bubbles nucleate uniformly over a continuous surface
important question in the development of more HA of many grain-boundary segments, then the rate of growth
resistant steels. Increasing the chromium content reduces of the sample will be given by the equations for Ebd, or Esd
the equilibrium methane pressure Pm, but also reduces the However, there will be some segments of grain boundary
alloy's creep resistance somewhat. A higher creep strength with no bubbles, and these regions will have to
can be obtained with addition of Ti or V (Ref. 54). accommodate the expanding boundaries on either side via
Parthasarathy49 argues that in the HA of commercial lattice creep. Steels apparently differ in the degree to which
Cr-Mo steels the fraction of boundaries with bubbles is the bubbled grain-boundary segments form a continuous
Downloaded by [Universite Laval] at 09:13 09 May 2016
sufficiently high for the cage restraint not to exist. Also, he path across the sample. This difference might be
obtains good agreement between his model and the represented by the ratio of the average values of IX and f3 in
observed data without including the restraint effect. It is Fig. 8e. If this ratio is 1 then the bubbled boundaries will
also noteworthy that there was a drop from 218 to 154 in form an essentially continuous path across the sample, and
the Vickers hardness between the two most highly the rate of growth will be given by Ebd' If the ratio is less
tempered samples in Fig. 5, yet the change in theHA rate than unity the initial expansion may be given by Ebd or Esd,
is negligible, that is, E is proportional to the activity of but after the bubbles on the bubbled segments grow to
carbon, but does not change with a change in matrix link up the next stage of growth will be controlled by
strength. lattice creep in the regions between the bubbled segments.
The values of Qm and m that should go into the
Experiment equations for E have not all been well determined for pure
Following the growth of submicrometre bubbles during the iron, but the best values are given in Tablel. They are
incubation period of HA requires either a very sensitive followed by one question mark if the value is based on
dilatometer (strain sensitivity less than 10- 5) or high- somewhat uncertain data, and two question. marks if the
magnification microscopy (over x 2000), Le. an SEM. It is values are educated guesses. It is possible that the grain-
only within the last few years that such techniques have boundary data are markedly influenced by the adsorption
been used to study HA. Work done previously could of impurities, as has been reported for nickel,56 but there
establish only the incubation time, that is, when the are no reliable data on the subject for steel.
bubbles link up and the expansion becomes rapid. McKimpson and Shewmon 19 measured the rate of HA
It has been demonstrated by dilatometry that at the of an ESR carbon steel (O'3%C) with a capacitance
hydrogen pressures of engineering interest (> 7 MPa) the dilatometer. They found appreciable change in the
bubble density quickly reaches a steady-state value, and the temperature and hydrogen-pressure coefficients, QH and u
growth rate t: becomes time independent, that is, E is a (equation (11)), on going from low Pm (high T, low PH) to
function only of temperature and hydrogen pressure during high Pm (low T, high PH)' In fact, QH increased from
the incubation period. 17.19 The results obtained can then 115 to 210 kJ moi -1, and the hydrogen-pressure exponent
be described by the expression u decreased from 19 to 0,6. This is what one would expect
t: = CP~exp(-QH/RT) (11)
from the non-ideality of methane (Fig. 4) if the same
process controlled the rate at both high and low Pm. If
where PH is the hydrogen pressure. these rate data are described in terms of Pm (equation (12)),
Since the methane pressure Pm is the actual driving force the Qm values for the two domains of Pm are essentially
that enters the mechanistic equations given above, clearer equal to 220 and 225 kJ mol- 1, while the values of mare
Table 1 Model predicted* and observed rate coefficients for hydrogen attack
Om(mod), m(obs),
Steel Control process kJ mol-1 m(mod) kJ mol-1 m(obs) A, }lm Ref.
* ? and?? after constants used in the model equations reflects the relative uncertainty in the experimental data available
for steel.
t Average of two BM heats.
15 and 12, respectively. The values of m and Qrn agree m = 1 in heavily tempered steel, and thus supporting grain-
satisfactorily with those expected for boundary-diffusion- boundary diffusion as the rate-limiting process, at least in
controlled growth, and Parthasarathy found that such a soft Cr-Mo steels.
model fitted the data wen over the entire range with a There are few data on the HA resistance of welds for QT
bubble spacing b = 05 Jlm and a bubble radius 2'25Cr-1Mo, and these results differ. Table 1 includes
a = 005 Ilm (Ref. 49). growth data on the HA kinetics of two commercial weld-
Panda and Shewmon45 used dilatometry to study an Al- metal samples (WM 1 - shielded metal arc and WM2 -
killed carbon steel that nucleated bubbles on an submerged arc).55 Both WM samples grew up to ten times
appreciably smaller fraction of the grain-boundary faster than the base metal (BM in Table 1). Erwin24
segments than McKimpson and Shewmon's, and indeed studied the HA of ESR welds and detected no difference
most steels that have been studied.1s The bubbles between the BM, WM, and HAZ. Sakai and Kaji57
nucleated on these boundary segments initially seemed to reported a bubble density orders of magnitude higher in
grow by surface-diffusion control until they linked up to the HAZ of a welded plate than in the BM. Masaoka et
form a single bubble over the grain-boundary segment. The al.27 coarsened the austenite grain size of 2'25Cr-1Mo
values of Qrn and m for this initial stage are given in steel, called it 'simulated HAZ' and reported essentially no
Table 1 under 'Surface diffusion'. In the second stage the change in HA kinetics at 600C. Clearly, the variability in
large bubbles covering isolated grain-boundary segments the HA of welds is greater than that found in BM, but
grew with Qrn and m values much larger than in the first beyond that there are too few data, or too little
stage. These data are given in Table 1 under 'Creep' as the understanding, to draw firm"conclusions.
controlling process. Effect of creep Modern PV design emphasizes the use of a
The obse~ved values of Qrn and m given in Table 1 for ductile material so that local stress concentrations will be
grain-boundary and surface-diffusion controlled bubble eliminated by localized plastic flow without introducing
growth agree reasonably with those of the model cracks. For PVs in which creep is unimportant room-
Downloaded by [Universite Laval] at 09:13 09 May 2016
predictions, but are consistently somewhat higher than the temperature elongation provides a good indication of
values expected. This could be explained by a small lattice- adequate ductility. However, with longer time service at
creep contribution from the regions between the individual higher temperature, creep becomes a factor and creep
bubbles on a bubbled boundary,49 or from creep ductility is necessary to relieve local stress concentrations.
accommodation of the unbubbled boundary segments Creep elongations of 015-020 are typical for QT Cr-Mo
shown in Fig. 8c. steels, the elongation being limited by the growth and link-
225Cr-1Mo steel To produce a loss of room-temperature up of voids on grain boundaries.
mechanical properties in QT 2'25Cr-1Mo steel requires In samples undergoing HA, failure (bubble link-up into
exposure to extreme conditions of temperature and fissures) can occur with all strain arising from the
pressure, for example, PH = 30 MPa at 600C for 1000 h or distribution over the boundary of all of the atoms that
more. However, if the steel is inadequately tempered, HA is were initially in the volume occupied by bubbles. In this
observed at lower temperatures and pressures. 55 case the strain at fracture is approximated by
The dilatometric study of HA in this steel requires
greater sensitivity than in carbon steel, but has been
err = bfb/2d . . . . . . . . . . . . . . (13)
successfully carried out. QT 2'25Cr-1Mo steels are similar where it has been assumed that link up occurs when a
to carbon steels in that e = f(T, PH) over the great majority grows to b/2. Typically for carbon steel b = 05 Ilm,
of the incubation period. However, the behaviour of the d = 20 Jlm, and fb = 01, or err = 0'001, which agrees well
two steels differs fundamentally in the initial (bubble with the observed value of strain at the end of the
nucleation) stage of attack. At 500-550C a freshly QT incubation period in carbon steel.19
Cr-Mo steel exposed to 20 MPa of hydrogen expands at a In creep studies one would thus expect a significant
steadily decreasing rate until after a few days e stabilizes (at reduction in err if HA played a role in developing the
550C) or becomes immeasurably small (at 500C). This boundary fissures. This is indeed found to be the case, as
initial strain reflects the build-up of methane in existing was described above in the section 'Observations'. The
voids in the Cr- Mo steel as opposed to the nucleation of surprising thing is how mild the hydrogen exposure can be
many new bubbles in carbon steels. This difference between and still have a significant effect on the creep ductility. In
the two types of steel is discussed further in the next unstressed samples exposed to PH = 20 MPa, one observes
section. a drop in room-temperature elongation after 1000 h only
Dilatometric data on QT 2'25Cr-1Mo steel plate are after exposure at 600C or above. However, in a
given in Table 1 (labelled 'BM' for base metal). The values stress-rupture test (O'n = 310 MPa) run in Pm = 20 MPa
of Qrn and m are comparable to those found in carbon steel hydrogen there was clear evidence of a hydrogen-induced
when boundary diffusion limits growth. The orders-of- loss in ductility on exposure at only 455C (failure in
magnitude slower growth in the Cr-Mo steels (at the same 4500 h).25 If hydrogen does not change the number of
temperature) can be explained as follows. For primarily voids/bubbles on the grain boundaries this would represent
boundary-diffusion-controlled growth the rate e is propor- an increase of about 100 in the rate of bubble growth. The
tional to (1/b2d)F1'7 The reduction in carbon activity gives effect of creep probably results both from an enhanced
a drop in Pm = F of about 3, which raised to the 17 power bubble growth rate, and an enhanced nucleation of
is a factor of 6. The bubble spacing along boundaries b is bubbles, but to date no studies have been made indicating
about 10 times larger in the Cr-Mo steel than in the the relative contributions of the two effects.
carbon steel, and the grain diameter d is about 25 times The enhancement of diffusion-controlled bubble growth
larger in the Cr- Mo steels. Multiplied together these by lattice creep would be entirely consistent with the
factors account for a difference in rate of 1500 between the theories of creep ductility in the literature. In the case
two types of steel. Alternatively, with a Qm of 225 kJ mol-1 of O'n = 0, the main force resisting bubble growth by
the Cr-Mo steel will attain the same e as the carbon steel grain-boundary diffusion stems from the requirement that
at 150 K higher temperature. However, since the bubbles the material leaving the bubbles diffuse and deposit
must grow further to merge in the Cr-Mo steel, 150 K uniformly over the entire region between the bubbles
understates the difference in useful temperature for the two (Fig. 8a). However, the strain rates observed in samples
steels. Figure 1 indicates an about 200 K difference. undergoing HA are only about 1% of that typical of
Figure 6 indicates a continuous drop in the force samples undergoing creep. Thus even relatively small rates
exponent m (equation (12)) with Pm' reaching a value of of creep due to an external stress could open up space
'
b diff-. -""*I-----I(reep~
11r----------j1
alloys, cold work resulted in an increased void density at
grain boundaries and a significantly reduced creep
ductility.
{=
::Z:1 z:::L??::z:? n? 9 b
b-a------, r-!
b-a---I t
Carbon steels
Pishko et al. 18found that the HA of carbon steels develops
a mean bubble spacing of 1 /lm on the grain boundaries
that form the fracture surface, but found no inclusions
9 Creep v. grain-boundary (gb) diffusion coupling
under influence of external stress Go: for Go = 0 entire
inside these bubbles after fracture. These bubbles form at
relaxation is by gb diffusion over a distance b- a, carbide/ferrite interfaces. Their density is much higher than
while with an applied stress of Go > 0, the stress in the density of any inclusions visible with the optical
Downloaded by [Universite Laval] at 09:13 09 May 2016
lattice normal to gb, cr(r), rises but is limited by creep microscope or SEM. Thus these fine bubbles must form in
in central region; reduction of gb diffusion distance a way that is fundamentally different from those found at
from b- a to b'- a allows matching of rates of the large (10-1000 /lm) inclusions in older dirtier steels.
diffusion- and creep-limited regions Using a two-stage replica technique, Wang65 showed
that. only a negligibly small density of voids existed in
annealed steel before hydrogen exposure. The development
along the boundary for the material leaving the bubbles, of a high density of bubbles on exposure to high-pressure
and thus substantially shorten the required diffusion hydrogen is shown by the kinetic studies of Panda and
distance (see Fig. 9). The effect of lattice creep on void Shewmon,45 who found that if a carbon steel was heated
growth under an externally applied stress has been treated under 20 MPa of hydrogen no growth occurred on heating
by several authors.50-53 These authors find that in an in stages (over 6 d) from 275 to 350C. However, at 375C
intermediate range of applied stress the rate of void growth growth started and in a few hours the rate increased by
by boundary diffusion is appreciably enhanced because of over two orders of magnitude. Following this surge of
the shortened diffusion distance for atoms leaving nucleation the bubble density did not change measurably
boundary voids, and the concentration of the applied load with time or further HA at various temperatures and
in the region b - b' of Fig. 9. This results in a higher creep hydrogen pressures. This stable array of nuclei allowed
rate around the plane of the boundary than in the matrix, growth to be measured subsequently down to 275C.
and though the sample expands by creep, the localization McKimpson and Shewm'on19 had earlier obtained similar
of the creep results in a low-ductility fracture. results and initially exposed samples briefly to a high
Wanagel et al.26 have shown that a pretreatment with hydrogen pressure (20 MPa) at 375-400C to establish a
13 MPa hydrogen at 600C significantly decreases the stable bubble density.
stress-rupture ductility of 2'25Cr-lMo subsequently tested At low hydrogen pressures the nucleation of bubbles
in air. Thus the effect of HA on creep ductility seems to be proceeds more slowly but is an essential first step to HA.
due both to enhanced nucleation and growth of the Wang65 has measured the initial growth of annealed
bubbles. samples at low hydrogen pressures (2 MPa) and
400-500C. She found the rate to rise in successive stages
NUCLEATION or steps from an initially undetectable level to an easily
As was seen above in the discussion of growth, the density observable one. Presumably, this was caused by the
of bubbles that nucleate is one of the most important spreading of the stable high-density (one per square
variables in determining the rate of HA of a steel. It is also micrometre) layer of bubbles over sheets of grain
the least well understood aspect of the mechanism of HA. boundaries across the entire cross-section of the samples.
A summary of the observations is given as a preliminary.
Chromium-molybdenum steels
Inclusions In contrast to carbon steels, when Cr-Mo steels are
The steels in common use up to a decade ago contained a exposed to high-pressure hydrogen (20 MPa), there is no
much larger volume of larger inclusions than is found in initial rise in the rate of growth that would reflect the
the structural steel currently being produced. Gas bubbles nucleation of additional bubbles. In fact at 500C an"
easily form at stringered sulphide or silicate inclusions of a 20 MPa of hydrogen the growth initially starts at an easily
size easily observed in the light microscope. 2,20,22,58This measurable rate and then over a period of a few days drops
is probably due to pre-existing voids or lack of cohesion at to a negligible rate.33 The indications all point to a pre-
these interfaces. existing set of voids that fill with methane on exposure to
hydrogen. Using a two-stage replica technique Lopez and
Cold work Shewmon66 have shown that in QT 2'25Cr-1Mo steel a
Cold working an annealed carbon steel greatly increases high density of voids is formed at the tips of carbides
the rate of HA. This leads to earlier fissuring along the during tempering, especially between grain-boundary
grain boundaries20 and an enhanced rate of growth for the carbides. The methane-bubble density that subsequently
decarburized surface layer.2,22 Shih showed that heavy gives rise to HA is much lower, and this density remains
cold work by swaging leads to the development of bands of essentially constant independent of time during hundreds
methane bubbles within pearlite colonies. 59 The 1970 of hours of exposure to hydrogen at 20 MPa (Ref. 67).
less continuous surface of bubbles through the sample. This the WM 1 and WM2 data shown in Table 1. It is seen that
leads to fissuring at relatively low sample strains (0001). It the weld-metal limits are clearly lower than for the base
clearly develops in carbon steels, and apparently develops metal, and that the effect of 100 MN m -2 static stress is
in QT 2'25Cr-1Mo steels at high pressures. It is postulated small.
that such active nucleation occurs at values of Such an approach ignores several factors; some are
F > 600 MPa in both carbon and low-alloy steels. helpful and some are not. It tends to underestimate the life
2. The formation of voids during the tempering of because:
2'25Cr-1 Mo, and the persistence of many of these through
(i) vessels have hydrogen on only one side and
long tempering operations to give rise to bubbles in HA.
hydrogen diffuses out the other side so that HA
This is the dominant process for F < 600 MPa.
proceeds only from one side. of the vessel
In carbon steel occasional voids are present at the ends (ii) vessels are usually clad with stainless steel and the
of carbide lamellae. The high density of bubbles nucleates requirement that the hydrogen diffuse through this
and spreads for Pm > 600 MPa and temperatures at or cladding to compensate for the hydrogen that
above 375C. The critical radius (2y/P m) for Pm = 600 MPa diffuses through the vessel lowers the hydrogen
is about 5 nm. If methane formed inside a void of this size activity in the PV steel beneath that for the PH in the
at say 400C it would force iron atoms into the boundary vessel.
tending to wedge open the interface at the next carbide
particle. If a void of > 25 nm were present there the This approach tends to overestimate the life because:
combined action of the methane inside the bubble and the (i) it neglects the accelerating effect of creep on the
wedging action of the atoms from the first void would bubble growth, and bubble nucleation
make it grow.41 These two bubbles would raise still further (ii) it neglects the local effect of residual stress and stress
the stress intensity in the surrounding boundary, and make concentrations around discontinuities which may
spreading of the bubbled region easier.69 Such a model be appreciably higher than 100 MN m - 2 in some
would require residual stress or small voids at the ends of regions
the carbides, but would allow the bursts of nucleation that (iii) there is no allowance for the scatter expected
are observed as well as the observed changes in the between different heats and different welds, and thus
morphology of bubble distribution.70 no lower bound on the variation to be expected .
besHindige Stahle'; Dusseldorf, Verlag Stahleisen. 46. T. J. CHUANG, K. J. KAGAWA, J. R. RICE and L. B. SILLS: Acta
7. S. HARPER, V. A. CALLCUT, D. W. TOWNSEND, and R. EBORALL: Metall., 1979, 27, 265-284.
J. Inst. Met., 1961-62, 90, 414-423, 423-429. 47. G. M. PHARR and w. D. NIX: Acta Metall., 1979,27, 1615-1631.
8. s. H. GOOD and w. D. NIX: Acta Metall., 1977,26,739. 48. B. F. DYSON: Met. Sci., 1976, 10, 349-353.
9. E. MITTEMEIJER et al.: Z. Metallkd., 1983, 74, 473-483. 49. T. A. PARTHASARATHY: unpublished work.
10. R. H. BRICKNELL and D. A. WOODFORD: Acta Metall., 1982, 30, 50. w. BEERE and M. V. SPEIGHT: Met. Sci., 1978, 12, 172-176.
257-264. 51. G. H. EDWARDS and M. F. ASHBY: Acta Metall., 1979, 27,
11. B. F. DYSON: Acta Metall., 1982, 30, 1639-1646. 1505-1518.
12. E. P. PARTRIDGE: J. Eng. Power (Trans. ASME), 1964, 86, 52. A. NEEDLEMAN and J. R. RICE: Acta Metall., 1980, 28,
311-324. 1315-1332.
13. R. J. RIOJA, M. YACAMAN, M. MORENO, and A. PERAZA: SCI'. 53. T.-L. SHAM and A. NEEDLEMAN: Acta Metall., 1983, 31,
Metall., 1982,16, 129-134. 919-926.
14. M. J. YACAMAN, T. A. PARTHASARATHY, and J. P. HIRTH: Metall. 54. P. G. SHEWMON, H LOPEZ, and T. A. PARTHASARATHY: in
Trans., 1984, 15A, 1485-1490. 'Research on chrome-moly steels', 1-8; 1984, New York,
15. L. C. WIENER: Corrosion, 1961,17, 137-143. American Society of Mechanical Engineers.
16. R. A. ALLEN, R. J. JANSEN, P. C. ROSENTHAL, and F. H. VITOVEC: 55. T. A. PARTHASARATHY, H. LOPEZ, and P. G. SHEWMON:
Proc. Am. Petrol. Inst., 1961, 41, 74-85. to be published in Me tall. Trans. A.
17. G. SUNDARARAJAN and P. G. SHEWMON: Me tall. Trans., 1980, 56. R. A. PADGETT and c. L. WHITE: SCI'. Metall., 1984, 18,
l1A, 509-516. 459-462.
18. R. PISHKO, M. G. McKIMPSON, and P. G. SHEWMON: Me tall. 57. T. SAKAI and H. KAJI: Tetsu-to-Hagane (J. Iron Steel Inst. Jpn),
Trans., 1979, lOA, 887-894. 1978, 64, 66-75.
19. M. G. McKIMPSON and P. G. SHEWMON: Me tall. Trans., 1981, 58. H. M. SHIH and H. H. JOHNSON: SCI'.Metall., 1977, 11, 151-154.
12A, 825-834. 59. H. M. SHIH: PhD thesis, Cornell University, Ithaca, NY, 1978.
20. M. RANSICK and P. G. SHEWMON: Metall . .Trans., 1981, 12A, 60. 'Steels for hydrogen service at elevated temperatures and
17-22. pressures', Publication 941, 1 edn; 1970, New York, American
21. G. A. NELSON: Proc. Am. Petrol.Inst., 1965,45, (3), 190-195. Petroleum Institute.
22. R. E. ALLEN, P. C. ROSENTHAL, R. J. JANSEN, and F. H. VITOVEC: 61. z. J. SU, D. STONE, J. WANAGEL, and CHE-YU LI: Int. J. Struct.
Proc. Am. Petrol. Inst., 1961, 41, (3), 74-85. Mech. Mater. Sci., in press.
23. G. SOREL and M. J. HUMPHRIES: Mater. Perform., Aug. 1978, 17, 62. s. F. CLUGSTON, J. R. WEERTMAN, and P. G. SHEWMON: Metall.
33-41. Trans., 1983, 14A, 695-699.
24. w. E. ERWIN: Proc. Am. Petrol. Inst., 1981, 61, (3), 120-134. 63. B. DYSON, M. S. LOVEDAY, and M. J. RODGERS: Proc. R. Soc.,
25. A. R. CUIFFREDA, N. B. HECKLER, and E. B. NORRIS: in 'Advanced 1976, A349, 245-259.
materials for pressure vessel service with hydrogen at high 64. T. SAEGUSE, M. UEMURA, and J. R. WEERTMAN: Metall. Trans.,
temperatures and pressures', (ed. M. Semchysen), 53; 1982, 1980, l1A, 1453-1458.
New York, American Society of Mechanical Engineers. 65. MING-MIE WANG: unpublished work, Ohio State University.
26. J. WANAGEL, T. J. HAKKARAINEN, and CHE-YU LI: in 66. H. LOPEZ and P. G. SHEWMON: Acta Metall., 1983, 31,
'Application of 225Cr-IMo steel for thick walled pressure 1945-1950.
vessels', STP 755, 93-108; 1982, Philadelphia, Pa, American 67. H. LOPEZ and P. G. SHEWMON: unpublished work.
Society for Testing and Materials. 68. J. R. THYGESON, Jr and M. MOLSTAD: J. Chem. Eng. Data, 1964,
27. I. MASAOKA et al.: in 'Current solutions to hydrogen problems 9, (2), 309-315.
in steel', (ed. C. Interante and G. Pressouyre), 242-248; 1982, 69. J. P. HIRTH: Res Mech. Lett., 1981, 1, 3-5.
Metals Park, Ohio, American Society for Metals. 70. P. G. SHEWMON, c. RICHIED, M. McKIMPSON, and M. RANSICK:
28. H. H. PODGURSKI: Trans. AIME, 1961, 221, 389-394. SCI'.Metall., 1980, 14, 545-548.
29. M. NATAN and H. H. JOHNSON: Metall. Trans., 1983, 14A, 71. w. E. ERWIN and J. G. KERR: Weld. Res. Counc. Bull., 1984,
963--971. (275).
30. T. A. PARTHASARATHY: PhD thesis, Ohio State University, 1983. 72. J. A. GULYA and R. A. SWIFT: in 'Research on chrome-moly
31. G. H. GEIGER and o. F. ANGELES: 'Study of effects of high- steels', 9-30; 1984, New York, American Society of
temperature, high-pressure hydrogen on low-alloy steels', Mechanical Engineers.
Publication 945; 1975, New York, American Petroleum 73. 'Research on chrome-moly steels'; 1984, New York,
Institute. American Society of Mechanical Engineers.
32. G. R. ODETTE and s. s. VAGARALI: Metall. Trans., 1982, 13A, 74. A. S. ARGON and J. 1M: Metall. Trans., 1975, 6A, 839-851.
299-302. 75. F. M. BEREMIN: Metall. Trans., 1981, 12A, 723-731.