You are on page 1of 17

A STUDY OF STRESS RELIEF CRACKING IN

2$Cr 1 MO STEEL-I. THE EFFECTS OF


P SEGREGATION
C A. HIPPSLEY’, J. F. KNO’TT and B. C. EDWARDS2
‘Department of Metallurgy and Materials Science, University of Cambridge, Pembroke Street, Cambridge
2Metallurgy Division. A.E.R.E. Harwell.

(Recciwd 2 1 Noamber 1979: irr rcctwd~/ornt I8 Decmber 1979)

Abstract-Stress relief cracking behaviour has been studied in controlled purity and (540ppm)
phosphorus-doped casts of 2$Cr 1 MO steel, using a test developed to simulate conditions in the
coarse-grained region of a weld heat-affected-zone (HAZ) during post-weld heat-treatment (PWHT).
Two distinct cracking regimes were observed in the phosphorusdoped alloy during the PWHT cycle.
At low temoeratures (700-750 Kk a smooth. intergranular failure mode was produced, whereas at higher
temperaturbs (800-875 K) the cracking mode changed to a ductile intergranular fracture, with voids
centred on MnS inclusions at the grain boundaries. The controlled purity alloy exhibited a similar type
of ductile intergranular cracking at the higher temperatures but 110 cracking was observed at lower
temperatures.
Auger electron spectroscopy (AES) revealed a maximum in the grain boundary concentration of
phosphorus during the stress-relief heating period of the phosphorus-doped alloy, confirming a predic-
tion based on segregation thermodynamics and kinetic theory. The temperature at which this peak
occurred was coincident with that of the low temperature cracking regime.

R&aum&Nous avons etude la fissuration au relLhement des contraintes dans des pieces coulCes en
acier 24 Cr I MO de purete contrBl& ou dope au phosphore (540 ppm), li l’aide dun essai permettant de
simuler la region a gros grains d’une zone affect&e la chaleur de soudure au tours dun traitement
thermique apres soudure.
Nous avons observe deux regimes dif&ents de fissuration dans l’alliage dope au phosphore, au tours
du traitement thermique aprts soudure. Aux basses temptratuces (700-750 K), on a un mode de rupture
deuce intergranulaire, alors qu’aux hautes temperatures (800-875 K), le mode de fissuration passait a
une rupture intergranulaire ductile, des cavites &ant centrces sur les inclusions intergranulaires de MnS.
L’alliage de pun& contr&e preset&tit un type analogue de fissuration intergranulaire ductile aux
temperatures tkvber, mais nous n’avons pas observe de fksuration aux basses temptratures.
La spectroscopic d’&ctrons Auger a permis de mettre en evidence un maximum dans la concen-
tration du phosphore intergranuiaire au tours du chauffage pour le rel3chement des contraintes dans
l’alliage do@ au phosphore, a qui confirme les previsions bas6es sur la thermodynamique et la theorie
cinttique de la tigregation. La temperature a laquelle ce pit apparaissait coincidait avec celle du rtgime
de fissuration de basse temperature.

Zusammenfassung-Die Verminderung von Spannungen durch RiBbildung wurde in gegossenem Stahl


2,25 Cr 1 MO von kontrollierter Reinheit und mit 540 ppm Phosphordotierung untersucht. Es wurde ein
Versuch durchgefiihrt, der die Bedingungen in dem grobkijrnigen Bereich eines durch die SchweiBtem-
peratur beeinflugten Gebietes wlhrend der Warmebehandlung nach der SchweiRung nachbildet.
In der phosphordotierten Legierung wurden wiihrend des Warmebehandlungszyklus twei unterschied-
liche RiBbildungsbereiche beobachtet. Bei niedrigen Temperaturen (700-750 K) entstand ein glatter
intergranularer Bruch, wohingegen bei hoheren Temperaturen (800-875 K) ein duktiler intergranularer
Bruch mit Hohlr8umen an den MnS-Einschliissen der Korngrenzen auftritt. Die Legierung mit kontrol-
lierter Reinheit zeigte bei hiiheren Temperaturen einen Phnlichen Typ duktiler intergranularer RiBbil-
dung bei niedrigeren Temperaturen wurde jedoch keine RiDbildung beobachtet.
Die Augerelektronen-Spektroskopie enthiillte in der phosphorhaltigen Legierung ein Maximum der
Phosphorkontentration in der Korngrenze, das wlhrend der Wlrmebehandlung zur Verminderung der
Spannungen auftrat. Dieser Befund bestPtigt eine Voraussage, die auf der Thermodynamik der Segrega-
tion und auf der kinetischen Theorie beruht. Die Temperatur, bei der dieses Maximum auftrat, stimmte
mit der des Rigbereiches bei niedrigen Temperaturen iiberein.

1. INTRODUCIION (PWHT) is therefore commonly employed to reduce


these residual stresses by thermally activated flow.
Residual stresses produced during the welding of This treatment has the additional benefit of decreas-
large steel components may increase the probability ing the hardness of the weldment, thereby increasing
of catastrophic failure and the rate of crack propaga- the toughness.
tion by fatigue, creep and stress corrosion mechan- However, the application of PWHT to low-alloy
isms during service life. A post weld heat-treatment steel components has caused several industrial fail-

869
870 HIPPSLEY, KNOTT ASI> EDWARDS: STRESS RELIEF CRACKING IN 2fCt I MO STEEL

ures [l-3] due to ‘stress-relief-cracking’ or ‘reheat tion theory [ 123and extrapolated diffusion data indi-
cracking’. This mode of embrittlement is characterised cate that appreciable P segregation may occur to the
by intergranular decohesion, either in a weld heat prior austenite grain boundaries during the heating
affected zone (HAZ) or in the weld metal itself. The period of the PWHT (see Appendix). Moreover, the
microcracks thus formed may either result in com- level of segregated P is predicted to maximisc well
plete fracture through the weldment or, more fre- below the stress relief ‘hold’ temperature (at a heating
quently, act as nuclei for subsequent crack propaga- rate of 1OOKh”) (Fig 1) decreasing to a relatively
tion during servix low value as the hold temperature is approached. On
Grain-boundary ductility may be reduced during this basis, the effect of P on seem-relief~acking
the stress relief cycle by grain boundary segregation of mechanismscould be much greater during the heating
trace impurities and/or by precipitation processes (e.g. period than at the hold temperature and therefore the
gram strengthening, precipitate free zone formation or initial investigation has concentrated on cracking be-
grain boundary carbide precipitation) to an extent haviour during the heating period only.
such that it is insufficient to accommodate the plastic
defo~tion associated with stress relaxation.
The susceptibility of alloy steels to stress relief 2 EXPERIMRNTAL PROCEDURE
embrittlement has been assessed using two main types
of test, in either of which, material may be obtained 2.1 Materials
from a weld HAZ or subjected to a thermal cycle to Alloys were prepared by vacuum melting Japanese
simulate HAZ microstructure: electrolytic iron together with high purity alloying
elements. The chemical analyses of the two alloys
(i) Constant load creep tests or constant strain
studied are presented in Table 1. The controlled pur-
stress relaxation tests at a tixed temperature [%7J,
ity alloy (A) has a lower content of residual elements
which monitor cracking only at the hold (i.e. maxi- than typical commercial casts of 24Cr 1 MO and the
mum) temperature of the stress relief cycle.
phosphorus doped alloy (B) is based on (A) but con-
(ii) Simulation of the complete thermal cycle pro-
tams an addition of 54Oppm P. The latter phos-
ducing conditions which more closely resemble a phorus level is greater than that found in typical com-
practical stress relief situation [g-l 13.
mercial steels but was chosen as a first step towards
The majority of this previous work has concen- understanding the segregation and precipitation
trated on deriving empirical relationships between phenomena in more complex commercial steel sys-
bulk chemical composition and stress relief cracking tems.
susceptibility for use in the selection of material for a Stress relief specimens were austenitised for 1 h at
given type of welded structure [3-S, 113. 1225K, quenched into oil and subsequently tempered
The present paper describes the first part of a study for 1 h at 925 K to produce a tempered m~tensite
which is designed to clarify the effects of specifsc im- parent metal PM) microstructure.
purities on the micromechanisms of stress-relief
cracking in a low alloy steel and hence develop a
2.2 Weat g&ted zone (HAZ) simulation
scientific basis for the empirical relationships. The im- The samples used in the stress relief cracking tests
purity element selected was phosphorus and the sys- were machined as notched 4 point bend bars (Fig. 2).
tem chosen was 24 Cr 1 MO steel, which is a weld&k However, these required much larger dimensions than
tow-alloy steel widely employed in the ekctricity- those of a typical coarse-grained RAE and it was
generating and petrochemical industries, decided to simulate the HA2 microstructure rather
Calculations based on current equilibrium segrega- than cut specimens out of a real weld. Emphasis was

hold

0 5 10 time thoursl
Fig. t. Stress relief thermal cycle. Heating rate and hold temperature are material and component
geometry specific.
Table 1. 240 f Mo steelscompo- in the weld meta{ (WMJ and decreasing with distance
sition (wt.%) into the parent metal (PM). Maximum residual
Cast No. A Cast No. B
stresses are thought to approach the PM yield
strength [ 173. During post-weld heat-treatment, the
Cr 2-25 2.25 surrounding material constrains the :veId area, giving
MO 0.97 1.02 8 boundary condition which approximates to that of
c 0.09 O.f2 constant totai displacement. Stress-relaxation takes
0.55 0.57
!? 0.025 0.02 place by plastic flow in the PM/HAZ/WM/HAZ/PM
Ni 0.02 0.03 combination, but the volume of the HAZ is much
CU 0.015 0.015 smaller than the volumes of the PM or the WM
Ti <O.Ol co.oi which lie within an appreciable residual stress field,
V <O*Of <O.Ol
W co.01 4.0%
Its ~nttibution to the overalt relief of stress is there-
Nb <o.ot <#Of fore negligible and, under constant displacement con-
S 0.006 0.006 ditions, the stresses experienced by the HA2 during
P 0.007 0.054 post weld heat treatment are dominated by the behav-
Sn <O.Ol co.01 iour of PMjWM areas.
Sb <O.Ol co.01
AS <O.Ol co.01 By loading the HAZ simulated, notched bend speci-
mens to various ioads in a 250 kN capacity MAND
ControIkd purity (A) and servo-hydraulic testing machine, ‘residual’ stress fields
P-doped (B) alloy compositions. of predetermined severity were obtained. The spcci-
mens were then heated at a rate of IOOKh-’ under
placed on reproducing realistic grain size and micro- conditions of constant total strain to a temperature of
structure, and a thermal cycle measured by 975 K at which point they were unloaded cooled and
Myers [l3f for the HAZ of a muttipass MMA butt broken open at 77 K. Since the HAZ area was only
weld in SCr )&IO )V steeI was used as a modeI for 7mm wide, macroscopic stress relief in the test was
simulation in the 2) Cr 1 Mo allow Some unnotched dominated by the PM and sensjtive monitoring of the
bend specimens were also testai. relaxing load indicated large scale cracking events by
Thespec&?nswereclampedbetweentwosetaof discontinuities of gradient in the otherwise smooth
water cooled copper blodcs in a rig similar to that of curve of load vs. temperature.
CiiRon and &orge[t4], the cent& area being air- Compensation for expansion of the specimeff/rig
rounded with an R.F. induction coil and concentrator assembly during heating was accomplished as follows:
leaving a uniform gap of 1 mm. Heating and cooling
rates were adjusted by variation of the R.F. power (i) a calibration was made of the ram displacement
and cooling block spacing, while the temperature was as a function of temperature required to maintain a
monitored with Pt/Pt-13% Rh wires spot welded di- constant, low load (-2 kN) on a dummy specimen
rectly on to the specimen surface. In this way a baini- over the full heating range;
tic miczosttucture was obtained in both alloys pas- Qif During an actual s~es~re~e~~a~king test at
sessing a mean grain diameter of 90~ similar to much higher initial goads (10 times that used for cali-
those reported for the coafso grained HA2 in 2$Cr bration) values of displacement appropriate to the tis-
1 MO weldments [l&16]. Hardness measurements ing temperature were fed back into the servo-
and optical microscopy showed that the m hydraulic command input so that the actual, moni-
grained microstructure was relatively uniform tored curve of ram displacement vs. temperature cor-
~rou~out the specimen thickness and to a distance responded to the recorded calibration within rt:PA.
of 3Smm on either side of the notch (I%_ 21. The
grain size varied within f loo/, of the mean for Details of the rig and control circuits are presented
material up to 10mm below the notch root. schematically in Fig. 3.
Five notched specimens of each alloy were tested at
2.3 Stress relief cracking tests
initial loads ranging from 3[y% to WA of the smooth
The residual stress field in a large weldment is bar PM general yield bad, as denoted by s&ces (a)
generally triaxial, the principal stresses being highest to (e) in Table 2. These ioads were equivalent to 25%
to t5% of the HAZ material notched bar general yield
load. Smooth 4 point bend specimens of both alloys
were also tested from an intermediate initial load. The
following observations were made:

(i) Inspection of bad vs. tern~ra~r~ curves for


gradient di~ontinuities indicating cracking events.
lhwtnonr tncm (ii} Measurement of crack lengths on specimens
Fig. 2. Stress relief cracking tcstpiccc design, showing fractured after a full heating period (or various stages
extent or simulated coarse gained HA2 region, thereof) to assess stress relief cracking severity.
872 HIPPSLEY, KNOTT ANDEDWARDS: STRESS RELIEF CRACKING IN 2tCr I MO STEEL

Fig 3. !&hematk diagram of test rig with servo-hydraulic controf and information record circuits.

Table 2. Initial load series with 2.4 Auger Electron Spectroscopy


corresponding notch opening dis-
placements (NOD) measured for Grain boundary rcgregation was monitored in the
P-do& alloy (B) P-doped alloy (B) at several stages throughout the
stress relief heating period using Auger Ekctron Spec-
NOD troscopy (AESI. Small AES impact sampies (5 mm di-
Load (kN) (x IO-* m)
ameter, 28mm long) were given a simulated HAZ
(a) te 0.33 thermal cycle by austenitising at 1573K and quench-
(b) 0.55 ing to form a partiaBy bainitic microstructure. The
(e) ; 0.78 samples were subsequently heated at 100K h- ‘ to a
r”f 2 a97 series of temperatures and water quenched to produce
e I.36
a set of specimens representati= of the stress relief
testpiece at the followingpoints in the heating period,
[iii) Examination of crack morphology by scanning 625 K; 725 K; 775 K; 825 K; 875 K; and 975 K.
electron microscopy (SEM) and optical metal- Samples were fractured in situ at - IO-t0 torr and
MaPhY* o*77 K using a VG i&ture stage. A hmi cyhndrica1
The crack surfaces were examined in the SBM first mirror amtlyser was used f= AES with an integral
as fractured, and subsequently after cleaning in a buf- electron gun providing a spot sixe on the specimen of
fered HCI sohttion to remove oxide. 4~. Differential energy spectra were recorded at a
Load IkN)

20 -

300 coo 500 600 700 800 900 1000 renptrrturc(lo

Fig. 4. Load vs. tem~rature relaxation curves for both alloys relaxed from initial Ioad 32 kN (condition
(C)l.
HIPPSLEY, KNOTT ANI) EDWARDS: !!XREiSS RELiEF CRACKING IN 2&Cr 1 MO STEEL 813

d@OQdl f ditQQtpQrQtUrQ) ~iSCOntlIwitlQS

hhal load lkNl


l m 0

40
8

8 8

8 8
20 -

a 3

760 800 960

hinperatute fKI

Fig, 5. Summary of cracking events detected in both alloys during the heating period for initial load
series (a-e). In the controlled purity ahoy (A) cracking was ohserved in the high temperature region (0)
whereas in the P doped alloy (B) cracking was observed at low temperatures f0) as well as in the high
tenqrature cracking regime fm).

modulation of 3 V peak to peak, a primary energy of with the gradient discontinuity in its foad vs. tempera-
2.5 keV and a qecimen current of 4 @. The acqui- ture curve.
sition and handling of the spectra were mana@ by a Cracking events identitkd by this method are sum-
Hewlett Packard 45 desk top computer with fioppy ma&d for the whole test series in Fii 5. There is a
disc storage [18]. significant difference in the behaviour of the two
alloys in that the controlled purity alloy crack5 in one
3. RESULTS temperature range only (8WK-875K) while the
Pdoped alloy crack5 in both this end a lower tem-
3.1 Stresss relief cracking tests perature range (700 K-750 K).
Unnotch~ 4 point bend sampks of both albys Figure 6 shows fracture surfaces after testing and
showed no indication of cracking after relaxation breaking open at 77 K. The stress relief cracks appear
from an initial load of 32 kN (condition (c) Tabk 2) dark becaus,eof oxidation. Bx and By are the speci-
and metallographic examination. The bad vs. tem- mens unloaded at points x and y in Fig 4 and ilhrs-
perature curves obtained for each alloy coincided, and trate crack growth over a 60 K/36 min period. Final
contained no sudden changes in relaxation rates, Le. stress relief crack kngths are presented in Fig. 7 as a
di~ontinuities in the bad vs. temperature curve function of initial loads For intermediate initial
gradient. loads, stress relief arcking severity, a5 de&mined by
However, such gradient diiontinuitie5 were surface sack kngth, is far greater in the P-doped
observed in the relaxation curves of notched spcci- alloy than in the controlled purity alloy. This difTer-
mens which also exhibited appreciable stress relief ence is not apparent for high and bw initial loads.
cracking. The curves of A(c) and B(c) are superim- Specimens relaxed from the lowest initial load con-
posed in Fig. 4. There is a discontinuity in the dition (a) exhibited no detectabfe load vs. temperature
gradient of the curve of R(c) at the point marked gradient di~ontinuit~ However, subsequent f&to-
‘crack initiation’. (Unfortunately, it has been necess- graphic examination revealed small areas of stress-
ary to reduce the dimensions of the original bad vs. relief-cracking.
temperature curves by a factor of 25 for presentation.
The gradient discontinuity marked in Fig 4 is not 3.2 Crack morphology
therefore as obvious as it is in the experimental curve 3.2.1, Fractography. Stress relief crack surfaces in ali
of B(c).) Two further specimens of B material were the controlled purity specimens consisted of heavily
relaxed from the same initiaf bad Ievet but unloaded cavitated intergranular facets. Some examples are
at the points marked x and y in Figure 4. Examin- presented in Fig. 8. All specimens except A(b) show a
ation of their fracture surfaces verified that the initia- uniform cavity size (52 pm) and distribution with
tion of cracking in the P-doped alloy, B(c), coincides small particles (-0.25 pm) visible within some of the
874 HIPPSLEY. KNOTT AW EDWARDS: STRESS RELIEF CRACKING IN 2fCr I Mo STEEL

(b)

A B BY Bx
(4 (c)
Fig. 6. Fracture surfaces of both alloys for initial load scria (a-c) and interrupted tests Bx and By.
HIPPSLEY, KNUTT ANDEDWARDS: STRESS REtlEF CRACKiNG IN Z&O I MO STEEL 875

cavities. These were identified as MI@ inclusions by ing can occur in a coarse grained HA2 rn~o~~~~
EDS microanalysis (Fig. 9). Sample A(b) contains a during the heating period of post weld heat treatment
high proportion of ductile tearing, and intergranular provided that the initial ‘residual’stress field is suffi-
facets with relatively large voids (Fig. 8). ciently severe. The ‘~urnbn~~ development of crack
Examination of the P-doped alloy series reveakd fronts (Fig. 6) suggests that stress triaxiality encour-
two crack morphalogies. The fracture surfaces of ages this type-of cracking, Crack initiation causes dis-
specimens B(a)-B(d) are dominated by ‘smooth’ inter- continuities in the gradients of load vs. temperature
granular facets, but also exhibit smaIi areas of heavily curves because the production of cracks acts as a
cavitated intcrgramdar f&e& at the stress relief sack second mode of stress r&if.
tips (Fig 10). However, it is the cavitated &et mor- The higher tempera&tre ctacrmg phenomenon
phology which dominates in specimen B(e), relaxed comma to both alloys is based on a cavitation meoh-
from the highest initial load. Cavities obsertred in the anism in which voids appear to nucleate on fine MnS
P-doped alloy series are of a size and distribution particles. Baker and Johnson [19] report a similar
similar to those of the c~rltr~lled purity alloy series in&granular facet morphology for fracture of an
and contain small particles, again ident%ed as MnS overheated NCMV steel containing 5Oppm S in
inchzsions. which voiding is centred on ~-MI& p&&s. An aver-
322 U#co~ ~~~. !%ct.$ons cut and isating mechanism may operate during the f&m&ion
polished perpendicular to the fraciurt surfa= of HA2 microstructure since this involves a s&B-
revealed secondary cracking only in specimens ciently high peak temperature (1645K) to diilve a
relaxed from the most severe initial strcas oonditions signi amount of sulphur which then precipitates
(A(e)and B(c)). Figure 11 illustrates extensive second- as MnS on prctkrcnthl grain boundary sites duria
ary cracking in sample B(e).The oavi&ted intergranu- cooring. In any evaluation of susceptibility to sfrcse
lar crack morpholo~ is ckar in Fig. 12 which also ~~f-~~k~~ it is therefm imporrant to employ a
shows an isolated area of secondary cracking in realistic HAZ-simulation thermal cycle since peak
sample A(e). Cavitated grain boundarii generally lie temperature and cooling rate are known to influence
normal to the direction of maximum principal applied the size and distribution of MnS precipitates in an
stress. overheated steel [i9].
MiddktonQO] has zrxzntly identified a grain
3.3 Auger Electron Specmseopy funny dispersion of MnS particles as an impor-
Within the stress relief heating period (295K- tant factor contribut~g to low ductility in bainitic
975Kf only specimens of the P-doped alloy (B) )Cr )Mo $V steels during hot tensile testing at
quenched from temperatures in the range 625 K- 973 K. He extends this behaviour to stress relief
875 K exhibited a suffi~~t umber of int~~~ul~ cracking, and discussesthe eff&ts of Sb, Sn, P and As
facets after fracture in the AI% system to enable re- impurities in terms of their idme 011grain bound-
liable grain boundary analysis. The elements present ary sulphide density rather than on cohesive strength.
in a typical spectrum (Figure 13) are P, S, MO, C, 0, In agreement with ~~dk~on’s findings, there is no
Cr and Fe. significant efkct of P on sulphide diiion and
Figure 14 shows the variation of peak height ratio hence on cavity distribution for the 24Cr 1 MO alloys
(as a percentage of the FelOl peak) with (stress-relief) examined, The dominant influence of P in the present
quenching temperature for P, Cr and MO.CL and MO stfcss relief cracking tests is the promotion of a separ-
do not appear to segregate signiicantly during the ate, more brittlt, cracking mechanism.
heating period_ but the grain boundary P eoncen- There is a strong correlation between the peak in
tration c&arty maxim&esal approximately 72s K. grain boundary P concentrations during the heating
period and the ‘lower-temperature’cracking phenom-
4. DISCUSSIQN enon which is associated with ‘smooth’ inkgranular
It is clear from these results that stress relief crack- fracture facets. Low-temperature stress-relief-cracking
876 HIPPSLI:1~. KSOTT ASI> EDWARDS: STRESS RELIEF CRACKING IN 2tCr I MO STEEL

A(c)

x 1250 x 6300

x 1200 x 6300

A(b)

x 1100 x 3000
Fig. 8. Scanning electron micrographs of controlled purity ailoy (A) stress relier crack surfaces.
HIPPSLEY, KNOTT AND EDWARDS: STRESS RELIEF CRACKING IN ZiCr 1 MO STEEL 877

0
<r, (X1

I I
I I I I I
0 1 2 3 4 5 6 I
keV

Fig. 9. Typiad X-ray ~~1~ ol a cavity inclusion (e.g.adjacent to ‘X’)and a cavitylip (e.g at ‘Y’&

coupkd with ‘smooW facets has been mported in temperature, high stress creep damage has been pro-
semal other steels [a, 8, II] but, as yet, there has posed by Debiez 181.This may be consistent with the
been no segregation data for compa&on other than type of fracture surface observnd, but cannot by itself
that for the ‘asquend& pm-test condition. account for the difference between the Pdoped aBoy,
The fractme ~~~ is similar to that found in
which would be held to exhibit wedge cracking, and
temper embrittkd alloy at& in which crack propa- the controlled purity alloy, which shows only cavita-
gation occurs in an unstable mode along grain boun- tion even under the most severe conditions of initial
daries whose cohesion has been k~wemdby segrega- stress. The low-temperature cracking process requires
tion of impurity elements. However, the low tempera- a combination of stress and temperature, on the one
ture stress-rehaf~acking involves relativeiy slow hand, likely to produce gram boundary sliding, and
crack caption rates (approx. 5 mm h-r in speci- on the other, to lower grain boundary cohesion by P
men B(c)) and if a brittk cracking mechanii were to segregation, promoting interface separation.
operate it would have to do so in a step-Iike manner, Auger electron spectroscopy confirms the qualitat-
advancing the crack front in discrete jumps. McMa- ive expectation of a peak in grain boundary P concen-
hon [ 1l] suggests that the mechanism may be one of tration during the heating period, as deduced from
cavity nucleation and growth and that the smooth the model described in the Appendix. Comparison of
appearance of intergranular facets is due to the heal- prediction and measturment aiso provides an insight
ing of cavities by self axon. Burns et ~2.@l] have into physical processes not induded in the model.
shown this to be possibk for cracks in the HAZ of a Chromium does not appear to segregate appreciably
Cr MO V steel weldment at 825K after ft3-1000h. during the heating period, whereas longer term
However, SEM and microanalysis revealed no tract embrittling treatments produce moderate grain
of MnS particks on the smooth intergranular facets, boundary Cr enrichment [22]. This is consistent with
which implies that, if cavities do form, they are not the stower ditfusion kinetics and larger ditfusion dis-
associated with the 8ne MnS ~s~~uti~s astance for Cr, compared with P, in a-Fe. A binary
observed at higher temperatures In particular, exam- F+P thermodynamic model is therefore more appro-
ination of specimen Bx, unloadcd and cooled from priate, for the time scak of stress relief treatments,
only 760 K, showed no evidence of voiding, (Fig. 15). than the ternary Fe-C-P model originally employed,
The diffusion rate at 76OK is relatively low, and it and is obtained simply by setting the Cr-P relative
therefore seems doubtful that a cavitation mechanism interaction coefficient a’$ to zero. However, neither
operates in the low-temperature cracking-range, Ieav- model takes into account the effect on P segregation
ing no detectable grain boundary deformation, of MO which is known to reduce temper embrittle-
whereas a similar mechanism in the higher range pro- ment in 2fCr 1 MO steel [23].
duces obvious cavities. The temperature of the predicted P peak is mainly
A wedge cracking process, characteristic of low a function of the free energy of segregation, AG”p,and
878 HIPPSLEY, KNOTT ASi> EDWARDS: STRESS RELIEF CRACKING IN Zfitr I Mo STEEL

x 700 B(C) cratktip x 450 B(el


Fig. 10. Scanning ckctron micrographs of P-doped alloy (B) stress relief crack surfaces.
HIPF‘SLEY, KNOTT AND EDWARDS: SfRESS RELIEF CRACKING IN 240 1 MO STEEL 879

x 120

Fig. 11. Secondary cracking in P-doped &lay (B) r&axed from most severe initial load (c).
880 HIIPPSLEY. KNGJ-T ASI) EDWARDS: STRESS RELIEF CRACKING IN 240 f MO STEEL

Fig 12. Seco~ld~y cracking in P-doped specimen B(c) showing grain boundary cavitation. Localised
secondary cracking in controlled purity specimen A(e)
HIPPSLEY, KNOTT AND EDWARDS: STRESS RELIEF CRACKING IN 2+Cr 1 MO STEEL 881

F81
WIDE

Inn 200 xx 400 600 coo


SV
AS FRACTURED

Fig. 13. AES spectrum taken from a grain boundary region in P-doped alloy with HA2 microstructure,
heated at 100 K h-’ and quenched from 725 K.

a value of 33.5 kJ mol-’ is required for coincidence The magnitude of the P peak varies with bulk P
with the measured peak temperature of 725 K concentrations (Xc,). If a Plza peak sensitivity factor
(Fig. 16). This value is lower than that used by of 2.07 [tfl is used to convert peak height ratio to
&ah [24] for predictions of time/temperature/embrit- grain boundary atomic percentage+an Xc, value of
tlement curves in a NiCrMn steel (43.9kJ mol-‘) and 180 atomic ppm is required for agreement between
also that of King and Wigmore [2!$] for a 3 Cr MO prediction and measurement (cf Fig. 17 vs. 14 rather
steel (4&OM mol-‘1 both steels being in the mar- than Fig. 16 vs. 14). The actual bulk P level is 975
ten&c condition. However, V~~~~ and Joshi atomic ppm and a reduction of this order in the
[26] have shown that grain boundary P segregation amount of P free to segregate is consistent with the
levels are generally lower in a bainitic microstructure reported scavenging action of MO on P in temper
than in a martensitic microstructure of CrMoV steel embrittlement studiu[23]. Molybdenum is thought
given the same embrittling heat treatment and that to interact strongly with P in bulk solution, restricting
the effect of microstructure.on segregation cannot be the fraction of bulk P which is free to diffuse to
explained in terms of kiitic factors This is consistent inteflaces.
with an apparently low AG”p in the bainitic 2)Cr Parameters, such as the &al stress relief crack
1 MO HAZ. length, or the initial load required to produce a given

Peak height ratto

-,03 10

1
8.4 t
*‘
,’
,’ -\

5 -
I

.,f..
i ‘,, o..;.
6

\
o..Cr

,\
f29

F-’ p120

4._6_.b._tio22,
‘b-‘-.-

0 f I . I t
300 LOO 500 600 700 800 900 1000 femprratutc {K)

Fig. 14. Variation of the P, Cr and MO peak height ratios (relative to Fe,& at the grain boundaries of
the P-doped alloy (B) as a function of temperature during the heating cycle.
882 HIPPSLEY. KNOTT ANI) EDWARDS: STRESS RELIEF CRACKING ‘IN 2iCr I MO S’TEEL

Fig. 15.Fracture surfaceofalloy B relaxedfrom initial load 32 kN (condition(c)) and unloaded at 760 K.
HIPPSLEY, KNOTT ANDEDWARDS: STRESS RELIEF CRACKING IN 2iCr I MO STEEL 883

ECWJBRlUM P SEGREGATION LEVELS plastic deformation occurred below the specimen


notch between load levels (d) and (e). A possible
explanation for the observed change in fracture
appearance from smooth intergranular facets to heavily
cavitated intergranular facets may lie in the
decohesion of grain boundary MnS/matrix interfaces
P during the plastic deformation caused by initial load-

I ing. The plastic zone may then contain many poten-


tial cavity nuclei which encourage secondary cracking
of the higher temperature type, and could inhibit the
lower temperature, P-induced cracking by either

300400 xxx 600. 700 eoosooma (i) crack blunting, or


TEMPmATlRE (K)
(ii) interruption of the segregated grain boundary P
layer by preferential segregation to the high energy,
fTEUCE0 G.B.P LEVELS decohered MnS/matrix interfaces.
A firm conclusion cannot be drawn from the
present evidenor and further work is obviously
required in this area.
No attempt has been made in this paper to assess
the effects of microstructural development during
post-weld heat-treatment on stress-relief-cracking
mechanisms. Consideration of empirical cracking sus-
ceptibility equations and current theories of temper-
and creep-embrittkment suggests that both types of
cracking observed may involve synergistic impurity/
microstructural mechanisms. Work is in progress to
TEMPERANRE 00

Fig. 16. Computed P segregation predictions (from Fe-P IO


EOUILlBRIUM P SEGREGAT!-OtdLEVELS
model). AC”, = 33.5W mol- ‘. 975 atomic ppm P.
0 sh

final stress relief crack surface area in notched, bend


specimens have been employed in the litera-
ture [9,1 I] to assess stress-relief-cracking severity or
susceptibility. McMahon et al. [l l] have also
reported sudden decrease s in stress relief crack surface
area with increasing initial notch opening displace-
ment (NOD)/Ioad which are consistent with the
results of this work presented in Fig 7.
It has been shown that the drop in final crack 800 900 1000
length in specimen B(e), contrary to the trend shown TEMPERATURE (K 1
by the other B-series specimens of increasing crack
length with increasing initial bad, is associated with a
change in fracture appearance from smooth to heavily
cavitated intergranular facets (Fig. lo), and extensive PliECiCTED G.B P LEVELS
secondary cracking (Fig 11). Only specimen A(e)
exhibited similar secondary cracking, and this was
limited to an isolated area shown in Fig 12. Thus the
total grain boundary area damaged by stress relief
cracking in specimen B(e) is far greater than measure-
ments of final crack kngth or crack surface ares
would suggest. This indicates that these parameters
alone may result in an unreliable assessment of stress
relief cracking severity or susceptibility without the
support of metallographic examination perpendicular
to the major crack surface.
It is interesting to consider the sudden change in 300 400 500 600 700 BCQ 900 1000

fracture mode in the B-series alloy with increasing TEMPERATURE IK)

initial load. The notch opening displacement (NOD) Fig. 17. Computed P segregation predictions (from Fe-P
values presented in Tabk 2 suggest that extensive model). AG”, = 33.5 kJ mol-‘, 180 atomic ppm P.
884 HIPPSLEY. KNOlT AKD EDWARDS: STRESS RELlEF CRACKING IN 2fCr 1 MO STEEL

investigate this aspect and also to study the stress- 4. J. Myers, Met. Technot. 5, 391 (1978).
relief-cracking behaviour of %-doped and commer- 5. P. Harris and K. E. Jones, Welding research retated to
power plant, Conf. proc. CEGB, Southampton (1972).
cial 2f Cr 1 MO alioys.
6. J. C. Ritter and R. McPherson, Met. Teciurof. 1, 506
Finally, it is clear that the detrimental effect of P on (1974).
stress relief cracking varies with the temperature at 7. R. 1. Presser and R. McPherson, Residual stresses in
which cracking occurs, being most severe in the tem- welded construction and their e&q Conf. proc. Weld-
perature range of maximum segregation. Thermodyn- ing Institute (1977).
8. S. Debiez, Residual stresses in welded construction and
amic and kinetic factors which determine the position their effects, Conf. proe. Wefding Institute (1977).
and ma~jtude of such impurity election peaks 9. A. D. Batte and M. C. Murphy, Mer. Technof. 6, 62
vary with impurity type (e.g. Sn, Sb, P) and therefore (1979).
the temperature ranges of their maximum influence 10. R. C. Miller and A. D. Batte, Met. Constr. 7, 550
(1975).
on stress relief cracking are unlikely to coincide. This
II. C. J. McMahon Jr., R. J. Dobbs and D. H. Gentner.
emphasises the importance of including the heating Mater. Sci. Engng 37, 179 (1978).
period in any assessment of the relative potency of a 12. M. Guttmann, Metali. 7kans. SA, 1383 (1977).
number of impurity e1ement.s. 13. J. Myers, CEGB infernal report R/M/R 161 (1972).
14. T. E. Clifton and M. J. George. _. Ner. Consrr. 1. 427
(1969)
5. CONCLUSIONS 15. E. Smith, 8. E, Blanchard and R. L. Apps, Welding of
creeu resistant steel. Conf. oroc. Wcldinn Institute
1. Two types of stress-reliefcracking can occur in (isjo), p. 79. . *
different temperature ranges during the heating 16. P. Alberry and W. Jones, Met. TecImol. 4.45 (1977).
period of a commercial PWHT, (i) a ‘higher tempera- 17. R. Fidler, CEGB tnternaI report R/M/R261 (3977).
18. H. E. Bishop. AERE R-9201 (1978k
ture’ form associated with ~~r~~u~ cavitation, (ii) 19. T. J. Baker -and R. Johnron, j.Z.S.i. 211, 783 (1973)
a ‘lower temperature’ form associated with ‘smooth’ 20. C. J. Middkton. CEGB inrernai reoort RD/L/R1995. ,
intergranular Facets. (1979). .
2. A fine, grain boundary dispersion of MnS pre- 21. D. Burns, D. W. James and H. Jones, Met. Sci. J. 7,
204 (1973).
cipitates is associated with the higher-temperature 22. T. Inoue. Grain boundnriea in en@reering mareriafs,
inter~~u~ cavitation found in both controlled- Proc. 4th Etohon Landing Coal, p. 553 (1974).
purity and P-doped alloys. The occurrence of this 23. Jin Yu and C. J. McMahon Jr., Metall. Trans. in press
type oF cracking in the controlled-purity alloy indi- (1980).
cates that stress-reliefcracking cannot be eliminated 24. M. P. !%ah, Acta MetaB. U, 345 (1977).
25. B. L. King and G. Wigmore, Metoll. Dens. 7Ail761
by restriction of bulk impurity levels while a grain (1976).
boundary dispersion of a suitable cavity nucleating 26. R. Viswanathan and A. Joshi, MetalL Trans. 6.4,2289
phase, such as MnS, remains. (1975)
3. The presence of P has the following effects: (i) A 27. B. C. Edwards, H. E. Bishop, J. C, Riviere and B. L.
Eyre, Acta Metall. Za, 957 (1976).
‘brittle’form of intergranuiar cracking is promoted in 28. M, Guttmann, Sud Sci. 53,213 (1975).
a temperature range coincident with maximum grain 29. D. McLean, Grain Boundaries fn Metals, Oxford Uni-
boundary P enrichment. (ii) The severity of stress versity Press (1957).
relief cracking, in terms of total grain boundary area 30. W. R. Tyson, Acta Metall. 26, 1471 (1978).
31. P. A. Stark, Introduction to nuwieoi methods (1970).
damaged, is increased. 32. P. Gruxin and V. Mural. Fix. Met. ~et~~~d 16. 551
4. Tests performed at or near the hold ~rn~~t~e (19631
alone are unlikely to give a true indication of the 33. P. Gruzin and V. Mural, Fiz. Met. Metalioued, 17, 62
relative efficacy of various impurity elements in 0964).
enhancing stress-relief-cracking. 34. Smithells, Metals Reference Book--&h edition.
35. E. D. Hondros and M. P. &ah, Int. Metall. Rev. 22,
262 (1977).
Acknowledgetnent~The authors would like to thank Pro-
fessor R. W. K. Honeycombe for provision of laboratory
facilities, members of the CEGB research hboratories at
Leatherhead and Southampton for vahable discussion,
and Mr. J. E. M. Braid for assistance in computer pro-
gramming. APPENDIX
Mr. C. A. Hippsley was supported by AERE Harwell
during the course of this work. Theoretical prediction of grain boundary segregation during
stress relief heating
fn a treatment of temper embrittk~nt in SAE 3140
REFERENCES steel. Seah 124) has combined equilibrium segregation
theory for a ternary Fe-M-I system[28] with diffusion
1. K. E. Jones. Weldina of creen resistant steels D. 66. kinetics [29] to predict time/temperature/embrittkment
Conf. proc. Weldinghtstitute i1970). . curves. This approach is extended here in an attempt to
2. L H. Taft and D. E. Ycldham. We/dine research related assess qualitatively the extent of grain boundary segrega-
to power plant, Conf. pr& CEG&, Southampton tion in a Fe&r-P system during the heating period of
(1972). PWHT. A computer program has been written which
3. 8. L. King. C. J. Middkton and R. D. Townsend, solves the relevant therrn~~a~ and kinetic equations
CEGB fnrcrnai report RDJJRi919 (1975). and integrates over the required thermat/tcmporal range.
HIPPSLEY, KNOTT AND EDWARDS: STRESS RELIEF CRACKING IN 2&Cr 1 MO STEEL 885

McLean 1291 derived the following expression for the i.j = M, I


kinetics of grain boundary segregation: AG’I = AG”i + z’ij Xh(m)j
Xh“i = saturation value of X/$m)i
AG“c = free energy of segregation of solute i in a binary
Xh(t) - X6(0) Fe-M system.
= I - exp(g)erfc($)“* (I)
Xb(m) - Xb(0) a’ij = relative interaction coefiicienl between cosegre-
gated solute species M and I, i.e. a’ij =
%<I - (%,I + %cM).
Xb(o) = grain boundary segregation level (gbsl) a( the
start of a period t of isothermal tempering. If il is assumed that segregation of both solute species
X6(t) = gbsl at the end of that period. can occur separately lo a maximum of one monolayer with
Xb(co) = gbsl after infinite time at the specific temperature no site competition. and that Xci, j < < I. equation (2)
T of isothermal tempering (i.e. the equilibrium may be rewritten [24]:
gbsl at 7’J. Xh(x)r
D = volume ditTusivity of the solute under consider- - Xci exp(AG’i/RT) (3)
I - Xh(z:)i
ation.
a = solute grain boundary enrichment ratio., Xh(r;)j
d = effective solute diffusion distance (i.e. grain = Xcj exp(AG’j/RT) (4)
I - Xh(c0)j
boundary width occupied by the segregated
solute). The heating period is approximated by a series of N
Tyson [30] has since modelled the kinetics for segrega- time/temperature steps. At/AT, and at each step tempera-
tion in a ternary system using a computer, and predicts ture, the equilibrium segregation levels Xhi,j(w)n are cdcu-
that grain boundary segregation generally occurs more lated using the modified Newton-Raphson method [31] for
rapidly than the McLean solution would suggest. However, the non-linear simultaneous equations (3) and (4). Equation
the discrepancy is only significant towards the later stages (I) then yields the following expression for segregation level
of segregation and the stress relief program employs the at the end of the (II + I)“’ step:
analytical, and therefore more easily manipulable, McLean Xhi(frl + 1) = [Xhi(l3)11 - X&(m)]
theory.
A simplified version of Guttmann’s [28] thermodynamic x [ 1 - exprs) erfc(F)“‘] + Xbi(tn) (5)
model for multi-component solutions gives the following
expression for the grain boundary equilibrium segregation
level of solute i at temperature T in a ternary Fe-M-I The program was run initially using data derived from
system: temper embrittlement studies on low alloy steels contain-
ing P or Cr. or both elements [24,25,32-351, and predicted
Xci exp(AG’i/RT) that an appreciable amount of P should indeed segregate
Xb(m)i = Xb”i L( (2) during heating. The model and parameters were then
1 + c Xcj[exp(AG’j/R’f) - I] reviewed in the light of experimental measurements using
J-1 Auger electron spectroscopy (cf. Discussion).

You might also like