GRAIN BOUNDARY SEGREGATION AND THE T-t

DEPENDENCE OF TEMPER BRITTLENESS

iliI. P. SEAH
Division of Chemical Standards. National Physical Laboratory, Teddington, England

Abstract-Low temperature brittleness, as measured by the ductile-brittle transition temperature

(DBTT). is a function of microstructure and grain boundary composition. At constant microstructure
the behaviour in temper brittleness is most clearly illustrated by the classic measurements of Cart,
Goldman. Jaffe and Buffurn in which a “nose” is observed in the temperature and time dependence
of the DBTT. Their measurements are compared in detail with the equivalent predictions for the
grain boundary segregation of phosphorus in their steel. The predictions are based on a slightly moditied
version of Guttmanhs equilibrium segregation theory for ternary systems and McLean’s analysis of
the kinetics of segregation. The correlation is very close and shows that these experimental measure-
ments are fully compatible with an equilibrium segregation theory without reference to the effects
of grain boundary carbides on segregation. The segregation predictions are confirmed by spot checks
using Auger electron spectroscopic measurement.
The calculations are then extended to other materials to determine the effect of the phosphorus
impurity and nickel alloy contents on the position of the temper brittle nose. Thus. conclusions may
be drawn relating to the minimum phosphorus purity level required in different alloys. for commercial
heat treatments, to avoid loss of properties associated with grain boundary segregarion.

R&urn&La fragiliti B basse tempirature, mesuree par la temperature de transition ductile-fragile

(TTDF). depend de la microstructure et de la composition du joint de grains. A microstructure con-
stante. la fragiliti- au revenu est illustree par les mesures classiques de Carr er al.. dans IesqueUes
on observe un “nez” dans la relation entre la TTDF, la temperature et le temps. On compare en
d2tail leurs mesures avcc Ies prtvisions de la segrigation intergramulaire du phosphore aux joints
de grains dans leur acier. Les predictions reposent sur une version I&&rement modifike de la thioriz
de la skgrkgation de Guttmann pour les systtmes ternaires et de l’anal>se de McLean sur la cinetique
de la stgrCgation. Les corrblations sont trts etroites et elles montrent que les resultats expCrimencaux
sont totalement compatibles avec une thkorie de la s&&gation g I’iquilibre. sans faire appel B I’sffet
de carbures intergramulaires sur la skgregation. On confirme les privisions sur la segrtgation par
des v&&cations ponctueiles par spectroscopic Auger.
On &tend les calculs B d’autres matiriaux pour determiner l’influencc de la teneur en limpureti
phosphore et de la concentration en nickel sur la position du nez de fragilitP au revenu. On tire
ainsi des conclusions sur la teneur maximale en phosphore to&able dans diffkrents a&ages avec lcs
traitements thermiques commerciaux. pour Cviter une diminution des propriPt&s lice B la s&grtgation
intergramulaire.

Zusammenfassung-Die TieftemperaturversprGdung, gemessen mit der Ubergangstemperatur duktil-

sprbd, ist eine Funktion von Mikrostruktur und Zusammensetzung der Korngrenzen. Bei konstanter
Mikrostruktur ist das Verhalten der AnlaDverspradung besonders klar belegt durch die klassischen
Messungen von C?rr, Goldman, Jaffe und Buffum, bei denen eine “Nae” in der Zeit- und Temperatur-
abhlngigkeit der Ubergangstemperatur duktil-sprljd beobachtet wird. Diese hiessungen werden im
einzelnen mit den gquivalenten Voraussagen fiir die Korngrenzensegregation von Phosphor in dsren
Stahl verglichen. Die Voraussa!en bauen auf einer leicht modifizierten Version der Gleichgewichtssege-
garionstheorie von Guttmann fur tern&e Systeme und der Analyse dcr Segregationskinetik von IMcLean
auf. Der Zusammenhang ist sehr eng und zeigt, daB diese Eaperimente vollstlndig mit der Gleich-
gewichtssegregationstheorie vertrHglich sind, ohne da!3 auf den EinfluD von Karbiden in den I;om-
grenzen auf die Segegation Bezug genommen werden mu% Die Voraussagen iiber die Segregation
werden mit punktf~~igen Proben mitteis spektroskopischen Augerelektrone-nm~sungen be&t&.
Die Berechnungen werden anschliel3end auf andere Materialien auseedehnt. urn den EinfluB van
Phosphor-Verunrkinigungen und vom Nickel-Legierungsgehah auf die L&e der i%se der AnlaOverspr&
dung zu bestimmen. So kijnnen Schliisse iiber das fiir die verschiedenen Legierungen erforderliche-
kleinste Niveau der Phosphorreinheit gezbgen werden, das notwaendig ist, urn bei kommerziellen
Wzrmebehandlungen den Verlust von Eigenschaften, die mit der Korngrenzsegregation zmam-
menhlngen. zu vermeiden.

*Here, as previously [3.X]. a monolayer of solute atoms 1. INTRODUCTION

is defined as a raft. one atom thick, of such packing that
each solute atom occupies the same volume‘as it would For many years there has been considerable contro-
in the compound in equilibrium with the solute at its maxi- versy about the cause of temper brittleness in low
mum solubilitv. This monolayer does not necessarily have
a physical significance, however, experiments show CZ5,26] alloy steels and much effort has been absorbed in
that it can be associated. quantitatively, with the mono- estabiishinp various aspects of the phenomenon. It
layers built up by grain boundary segregants. now seems agreed that temper brittleness is promoted
345
346 SEAH: GRAIN BOUNDARY SEGREGATION

by the grain boundary segregation of certain residual brittle transition temperature (DBTT), of a certain
impurities in steels [l-3] although it is not yet agreed heat of SAE 3140 steel of composition: (98 by mass):
precisely how the segregation occurs. This lack of
agreement may be because the field still lacks, to C Mn Si S P Ni Cr MO V N
0.39 0.79 0.30 0.028 0.015 1.26 0.77 0.02 <O.Ol 0.006
some extent, a proper definition of temper brittleness.
In this paper the term will be defined later by the The material was commercial quality bar stock, aus-
behaviour described below which is illustrated by the tenitised for 1 hr at 9OOC, water quenched to give
classic results shown later in Fig. 2. Briefly it concerns 100% martensite with an austenite grain size of
the increased propensity of low alloy steels to brittle ASTM 8 (22,~m). then tempered for 1 hr at 67X
intergranular failure at temperatures below 200°C due and again water quenched. This heat treatment was
to heating for prolonged times in the temperature known to give a low DBTT of -83°C and so this
range 325-575’C. This problem is generally associ- became their unembrittled reference state. The mater-
ated with the manufacture of large components for ial was then given isothermal heat treatments for
which the time in the above temperature range is between 10 min and 200 hi at given temperatures in
necessarily long. the range 375-650°C. The shifts in DBTT from
It is important to understand the mechanism of the reference state gave the measure of embrittlement
temper brittleness in some detail so that predictions and, when plotted as constant brittleness curves as
for low alloy steels may be made by extending the a function of time at temperature, gave the Tt
very few existing experimental observations into dif- embrittlement diagram shown in Fig. 1.
ferent alloys, longer times at temperature, different In a more detailed study of the double nose appar-
levels of purity and to different embrittling impurities. ent in Fig. 1, Buffum and Jaffe [15] discovered that
This understanding is not. of course, just relevant to the DBTT in the embrittled state was microstructure
temper brittleness since temper brittleness and grain dependent to the extent of 2°C per Rockwell “c”
boundary segregation have also been related to the hardness number increase, i.e. the DBTT rose as the
wider problems of creep embrittlement [&a, poor material became harder. The DBTT’s shown in Fig.
fatigue resistance [8,9]. intergranular stress corro- 1 were then corrected to constant hardness and the
sion [2, IO-121 and intergranular corrosion [13]. It is new Tt diagram, as depicted in figure 2 taken from
the intention of this work to show how one particular Cat-r, Goldman, JalTe and Buffum’s [16] paper, exhi-
mechanism of segregation. namely equilibrium segre- bited only one nose. This Tt embrittlement diagram
gation, provides a very accurate description for the represents that component of temper brittleness for
time and temperature dependence of the fullest set constant microstructure and grain size and is the be-
of temper brittleness data currently available. haviour understood by many, and used by many in
Thus, the best starting point to understand temper reviews [I, 17-193, to illustrate temper brittleness. The
brittleness is the set of experimental results, obtained embrittlement in this behaviour is usually termed
for the effects of time and temperature of heat treat- “reversible” since the steel, reheated to a temperature
ment on the degree of embrittlement of SAE 3140 just above the nose and then quenched, will be fully
steel, by the group at Watertown Arsenal Laboratory de-embrittled even with reheat times as short as one
in the years between 1948 and 1953. In 1950, Jaffe minute [20]. After de-embrittlement the material may
and Buffum [14] of that laboratory, using standard then be re-embrittled as shown in Fig. 2 but with
Charpy V notch specimens, evaluated the degree of times to any given level of embrittlement being some-
embrittlement, as determined by the shift in ductile- what shorter than in the first embrittlement tests [21]

xa I I I
O-l 1 100
Embrittling time. hS

Fig. 1. The time-temperature diagram for embrittlement of SAE 3140 after JafTe and Buffum[14].
 SEhH: GRAIN BOUNDARY SEGREGATION 347

uniess the de-embrittlement treatment is as long as

the original temper treatment. RELATION TO THE CARR cr a.!. RESL-LTS [16]
The above phenomena may be fully described by
2.1 Defining the emhrirtling specks
the kinetic approach to equilibrium grain boundary
segregation for certain residual impurity elements in Because a full analysis of the Carr er al. [16] material
the steel. This idea was first proposed generally by has not been published and Auger electron spec-
McLean and Northcott [22] and has gained much troscopy [.4ES] was not availabt at that time to
support over recent years. However, it was not until measure the grain boundary segregants, the relative
very recently that a theory, with suitable simplifying effects of the possible embrittling impurity elements
approximations, became available with the full poten- must first be estimated. A typical analysis by AES
tiai to describe the above results in detail. This of an inter~ranu1ar failure in a commerciai 33 NiCr-
theory [23], with a few minor m~i~cations to allow MoV(mass 9,) (0.26C, X40%, l.%Cr. 0.48X10, O.lOV)
for experimental observations, is evaluated numeri- rotor steel after the step cooled treatment devised by
cally in the sections that follow to describe Carr, Low. Stein, Turkalo and Laforce [27] (1 hr 593’C.
Goldman, JaKe and Buffum’s results [16]. In the past, 15 hr 538’C, 31 hr 524’C, 48 hr J96C. 72 hr 168’C.
theories have been proposed which account for equi- cool to 315’C and air cool), is shown in Fig. 3. The
librium grain boundary segregation in binary sys- steel had residual impurity levels of P, 90; Sn, 70:
tems [21.25], however, Guttmann [23] has extended Sb, 70; As. 25Owt ppm. This spectrum shows that
the analysis to equilibrium segregation in ternary sys- the grain boundaries in the material contain
tems in which the co-segregating species have a 0.25 monolayers of P, only 0.035 monolayers of Sn,
simple interaction term. It is this interaction term less than 0.07 monolayers of As and less than 0.01
which gives rise to the much higher segregation pro- monolayers of Sb and most other impurities. It is
pensity of species such as P, Sn and Sb in alloy steeis likely that the Sn level in the steel used by Carr,
in the embrittlement region beiow 525°C compared Goldman, Jaffe and Buffurn [16] was lower than that
with pure iron [26] and which also thus gives rise of this recent 3$ NiCrMoV steel as the higher tin
to the rapid de-embrittiement at temperatures above levels have been brought about, over recent years,
575’C when the segregation returns to values ap- by the increased use of scrap recycling in steelmaking.
proximating those in pure iron. The mechanism of An indicator of the possible purity level is given in
how the grain boundary segregation promotes brittle a later work of Carr et al. [ZS] on the same type
grain boundary failure is a problem which can be of steel as the earlier work and in which the Sn, Sb
separated completely and is dealt with elsewhere [3]. and As levels were only 12, 8.and 110 wt. ppm re-
In the sections that follow, the basic approach to spectively. Thus, it is probably quite accurate to
the results of Carr er nl. [16] is explained and Gutt- assume that all of the embrittlement in the steel used
mann’s [23] theory, with minor modifications, is eva- by Carr et al. [t6] was due to P segregation.
luated for the time and temperature dependence of Separate studies in this laboratory show [2] that
segregation. The results are then compared in detail intergranular failure in many commercial alloys is
with the me~urements shown in Fig. 2. Finaliy, pre- dominated by P segregation and that, of the other
dictions are made concerning the effects of changing possible segregants. only Sn is detected. Analysis of
the purity level and the alloy content on the Tt plots the temper embrittled steam turbine disc involved in
of embrittlement in steels, the failure at Hinkley Point in 1969 [29] shows that

0.1 10 IW 5GP
Embrittling time~Ohrr

Fig. 2. The time-temperature diagram for embrittlement of SAE 3110, corrected to constant hardness.
after Carr, Goldman, J&e and Buffum [16].
343 SE&f: GR.-ZIN BOUNDARY SEGREG.~TIC@i

Fig. 3. Auger electron spectrum from the intergranular fracture surface of a commercial 3fNiCrMoV
rotor steel.

P again is the main culprit [l]. Comparison with the terface 1311 or the carbide can displace an alloying
DBTT data for this steel [29] shows that, if a linear element which similarly, because ol’ coupling to the
relation exists between the P segregation level and impurity elements, causes enhancement of the impur-
the DBTT, the tatter shifts some 300°C per monolayer ity element concentration at the c~bide~matrix inter-
of P segregated. Such a linearity has often been ap- face or at the grain boundary near the carbide [30].
proximatety demonstrated but is seen clearly in the In the former theory the carbides have to grow to
recent work of Ohtani ef al. [30]. If the above DBTT quite large sizes to produce a measurable effect, how-
shift per monolayer of P segregated is applied to the ever. in many temper brittle steels transmission elec-
Carr er al. Cl63 results, levels of segregation in the tron microscope measurements [32] show that the
range 0.06 to 0.40 monolayers of P are required to grain boundary carbide thicknesses are two orders
describe the measurements in Fig. 2. This range of of magnitude too low for this effect to be significant,
levels is just that appropriate to an analysis in terms being generally in the range 3%1OOnm. At these
of the kinetic approach to equilibrium segregation. thicknesses the impurity enrichment by carbide rejec-
In the section that Follows the equilibrium segrega- tion is negligibie. In the second theory, Ohtani et
tion of P at the grain boundaries in the steel used ai. [SO] suggest that the growing carbide rejects Ni
by Carr et al. [16] will first be evaluated. Kinetic which accumulates at the grain boundary and,
diffusion terms will then be included to evaluate how because of the ?&impurity element coupling, impur-
the P segregation level increases with time at any ity atoms are drawn into the grain boundary. Both
given isothermal embrittling temperature. Thus, a Tt of the above effects of carbide rejection can occur
plot for P segregation may be fully evaluated numeri- but, in practice. may not be generahy important and
cally and compared with the equivalent plot for in the following analysis of Carr, Goldman, Jaffe and
embrittlement in the work of Carr et al. C163. Buffurn’s [16] rest&s, will be ignored. The effect is
transient and, as the rate of carbide growth falls, even-
2.2 Choice of adsorption theory tually disappears. Viswanathan and Sherlock [33]
Many theories for segregation have been proposed have shown that the embrittlement in commercial
in the literature. The main difference, at present, lies NiCrMoV steels. where the yield stress is unaffected
between the kinetically limited equilibrium segrega- by overageing. increases monotonicaily during heat
tion theories [23-251 and the theories which involve treatments as long as one year. No transient effects,
enhanced temporary segregation due to the growth which could be associated with carbide rejection, are
of carbides at the grain boundaries [30,313. In this observed. A similar monotonic increase can be seen
paper attention is concentrated on the former as in figure 2 and this precise increase is fully explained
being relevant to the type of behaviour as exemplified below in terms of the equilibrium adsorption theory.
by the measurements of Carr et al. [16] but some
2.3 Ecaluation of the equilibrium segregation in a ter-
words of comment must be made about the other
nary system
approaches. In theories involving the growth of car-
bides to enhance segregation two difIerent effects are Several theories of equilibrium segregation have
proposed, but both, unlike the kinetically limited been proposed, some applicable to binary S~S-
equilibrium segregation theories, give transient effects terns [24,25-J and some to more complex sys-
which eventually disappear in long term isothermal tems [23,25]. A11are basically the same but highlight
heat treatments. In these theories, either the growing different aspects of the general theory. For the anaiy-
carbide can displace impurities which, because of’ sis of temper brittleness, the theory recently proposed
limited diffusion. accumulate at the carbide:matrix in- by Guttmann [33] allows the easiest direct descrip-
 SEAH: GRAIN BOUXDARY SEGREGATION 349

tion. From Gurtmann’s equation (11) the level of less than one monolayer. At high temperatures the
segregation in a ternary system is given by, after cer- impurity behaviour in Guttmann’s theory is similar
tain approximations: to that in binary systems but, as the temperature falls,
Y
i bi Xci exp(AGI ‘R?“) the segregation levels of the impurity and alloy ele-
yo= 41 i, j = I, iM ments rise. causing the free energies of segregation
‘-bi
1 + C X,j[exp(AG; RT) - I] of each other to rise so that at some particular rem-
,?I
(1) perature a local condensation occurs and most of the
where X,, and Xbncare the grain boundary mole frac- grain boundary sites become filled. The real point of
tionai monolayer segregation levels of the impurity departure from previous theories lies in the analytic
and alloy elements respectively with bulk mole frac- expressions in equations (2).
tions of X,, and X,t. XFi is simply the saturation value In order to predict the level of segregation in SAE
of Xbi. AG; and AG;r are the free energies of segrega- 3140, values of AC,, AG, and rin are needed and,
tion of the two species to the grain boundary and if the coupling is strong, a measure is required for
depend on their mutual segregation levels according the relative effects of the ahoy levels (mass “,,)
to: 0.79 Mn, 1.16 Ni, 0.77 Cr, etc., present in the steel. To
AC; = AGr t &,Xbh, obtain some judgement of the relative magnitude of
(4 these parameters, the literature has been searched for
AG;, = A%, + I;,,&+, 1.
all measurements of grain boundary segregation of
where AG, and AG4r are the free energies of segrega- these elements in association with P in iron and steels.
tion of the impurity and alloy elements in their res- The results are shown in Table 1 [26]. The heat treat-
pective binary systems with iron and &t is the relative ment cited there is the last appropriate heating and
interaction coefficient between the co-segregating spe- S.C. represents the step cooled treatment of Low et
cies equal to the difference in the interaction coeffi- al. [37] cited earlier. The values of X, given are eva-
cients xsll and zF*r i rFcM.These coefficients, in turn, iuated directly from the published Auger eiectron
are merely the changes in nearest neighbour molar spectra and quantified by procedures established at
bond enthaipies upon forming the appropriate M-I, the NPL, examples of which are given by Seah and
Fe-I and Fe-M bonds. Hondros [25]. These figures may differ by up to an
The above equation (1) from Guttmann’s [23] order of magnitude from those given by individual
paper may be derived in a similar way to authors in the references cited in Table 1 because the
McLean’s [N] theory for grain boundary segregation effect of Auger electron ranges and backscattering
in which there is a fixed. limited number of grain were often ignored. Pb is just the grain boundary
boundary adsorption sites which may or may not be enrichment factor defined as the ratio X,:X,.

Table 1. Measurements of grain boundary segregation of phosphorus, nickel and chromium in low alloy steeis

0.X IKi 0.02P S.C. P 0.79 !I80 0-t)

\ii 0.39 11
OK L’cr 0.06P SC. P 0.58 536 (35)
CC 0.23 13
0.K 3 5x1 0.06P S.C. P 0.34 314 135)
Xi 0.17 3
0.K 1.7Cr OSXP S.C. P O.&L 590 135)
Cr 0.35 19
Ni 0.66 20
O.?P ICQO'C
WQ P 0.48 134 136)
LCCWC FC P I.15 319
I hr 700X WQ P 1.20 333
O.OYP I hr 7oo’C FC 37% WQ P 0.33 302 136)
0.05P 1 hr 7CO’C FC 375’C WQ P 0.19 211 136)
I hr 7W’C FC 375’C WQ P 0 IS 201
3.5% I.612 0.05P SC. P 0.71 ?8S (37)
Ni 047 14
SC. P OX 907 137)
Ni 0.95 19
0.4c t.xr 0.05P SC. P a.17 303 f37)
0.1X 3 07Cr 0.01P 5 hr XO’C P 0.43 II92 (38)
Cr 0.40 il
0. m 2.YKr O.OZlP 5 hr 55O’C P 0.49 1132 138)
o.oYc Z.‘)Z.Mn O.O?P 5 hr 55O’C P 0.31 860 (N
0.1X O.?3Si l.jMn 0.016P 5 hr 550°C P 0 15 867 (381
0.1X ?.98Cr 0.024P Au% WQ P I
0. 3 300 (38)
OCQC 2.934n O.O?P 5 hr 5M)‘C P 0.23 635 (38)
02°C O.O?SP 5 hr 5CO’C P 005 99 i?SI
Cast Iron 0.07OP A4 ret P O-IS 380 (3)
Mild Steel O.O??P As ret P 0.20 277 (2)
En 24 0.016P AS ret P 0.08 134 (?I
ICMOV 0.022P As xc P 0.20 482 (11
0.03P .AS ret P 045 831 (1)
O.@XP .AE ret P 0.21 11&t (2)
0.009P S.C. P 0.24 Ii80 This work
Xi 0.37 11 Fi$ 3
350 SEAH: GRAIN BOUNDARY SEGREGATION

The relative effects of the alloy &MeIltj &lo. Cr, Table 1 indicates that. at the grain boundary, the
Si. Mn and Ni can be assessed as follows. In commer- ratio of the excess Ni to P levels is 0.72. The matrix
cial steels both MO and Cr will, to a large extent, Ni at the grain boundary can also interact with the
be removed from solution by the formation of car- segregated P so that the effective Ni to P ratio rises
bides of the type MIC and MzjC6 leaving excess car- to 0.8 + 0.2. From Fig. 4, this approach gives a value
bon to form cementite. Table 1 shows that if the Cr for r& of 4.0 + 3.0 kcal mol- ‘, in the same region
level is higher than that necessary to use up all of as that shown in Table 2. Very recently Gutt-
the carbon. the excess Cr may couple to the P and mann [41] has re-evaluated the interaction coeffi-
then give high P grain boundary enrichment whereas cients z& of various alloy elements with phosphorus
for lower Cr this is not so. Thus. in SAE 31-W both in iron. The new values are based on Kaneko et
MO and Cr will be ignored. The~odynamic al. [42] experimental measurements of the effect of
tables [39] show that both Si and Mn form weaker these alloy elements on the solubility of phosphorus
bonds with P than does Fe. Thus. it is likely that, in iron and are more appropriate to segregation than
in SAE 3140 only the Ni-P coupling need be con- the earlier values. These measurements give a value
sidered. Table 1 shows that this coupling is, in gen- of 7.2 k 1.0 kcal mol-i for Z;ip* in approximate
eral, very significant. The free energy of segregation agreement with the experimental value in table 2 and
of Ni in pure Fe, AGsi, cannot be obtained directly the above approach.
but may be determined via the relation of fi with A further point to note from Table 1 is that, whilst
the inverse of solid solubility [25]. The values of the the Ni and P segregation levels separately are always
free energies of segregation giving the best description below one monolayer, their sum generally exceeds
of the results shown in Table 1 are shown in Table this limit. Simifar observations occur in studies of
2. In fact all of these free energies should have tem- segregation in iron in these laboratories Et.51 where
perature dependent terms because the entropy factors the total segregation of Sn and S can exceed one
in grain boundary segregation are quite signifi- monolayer. In this case there is no signi~cant compe-
cant [40]. However, the increased complexity only tition for sites between these species and, again,
obscures the arguments that follow and do not mater- Ramasubramanian and Stein’s [36] results show that
ially add to the discussion since the results need only 0.4 monolayers of S do not affect the level of P segre-
be evaluated over the limited temperature range 350 gation. To accommodate these observations, equation
to 600°C. (I) may be rewritten in the approximation of a full
The value, 6.2, of &. derived from the experimen- monolayer of possible sites being available for the
tal measurements of Table 1 and shown in Table 2 segregation of both P and Ni separately without com-
is much lower than the 18.0 originally predicted petition for sites by the co-segregating species:
theoretically by Guttmann [23]. This may be because X,, exp(AG;/RT)
the relative amounts of Ni and P at the grain bound-
” = 1 + X,,[exp(AG,/RT) - I]
ary are such that NiP is a more appropriate “com-
pound” from which to evaluate @hiprather than the or. rearranged, since Xe 4 1
Ni,P used by Guttmann. & may be re-evaluated
XP
roughly as follows, based on Guttmann’s approach. ~ = X, exp(AG,/RT).
1 - xp
zNipand zFepmay be calculated for all Ni-P and Fe-P
compounds, as shown in Fig. 4, from the enthalpy Equation (3) is simpler than Guttmann’s version and
of formation given by Kubaschewski et a/. [39]. The leads to much easier calculation.
difference between these values, since XFeSiis small, A similar equation to (3) may be written for Ni
gives ahip which falls from Guttmann’s original value segregation also limited to one monolayer segrega-
of 18 kcal mol-*, if the coupling is such as to make tion. Observations of Ni segregation in the grain
the Ni and P form the compound N&P at the grain boundary zone by Joshi and Stein [43] and many
boundary, to near zero, if NiPL is more appropriate. others have shown, by combined Auger electron spec-
The constraints on structure imposed by the grains troscopy and ion etching that, unlike all other ele-
either side of the boundary make it unlikely that any ments, Ni is segregated to the grain boundary with
particular stoichiomet~c compound structure is a concentration profile extending at times up to 15
IocalIy formed around any segregant atom. so it layers either side of the boundary. This seems to be
seems most reasonable to choose a value for z& from fairly typical of all the Auger electron spectroscopic
Fig, 4 which is appropriate to the reiative levels of measurements of Ni segregation although, it must be
Ni and P found experimentally at grain boundaries added, that all the measurements happen to be for
in alloy steels. An average of the results shown in one, AISI 3340 (0.4C, 3.5Ni, 1.7Cr), rotor steel. If the
Table 2. Free energies of segregation in iron and Ni alloy observed Ni in the first atom layers either side of
Steels in kcal mol-’ the grain boundary is quantified, using the NPL pro-
cedure as in table 1, there is always less than one
Pure Iron Ni Alloy Sreel
monolayer. The excess Ni atoms further from the
P JGp = IO.5 i\Cp = 10.5 + 4.2 X,, boundary are unlikely to have a significant effect on
Hi bG,, = 2.75 AGG;,= 175 + 6.2 x,
the P segregation level as the coupling will be
 SEhH: GRAIN BOljNDARY SEGREG.ATION

Fig. 4. Values of the interaction coefficients, xMRbetween P and Ni or Fe for several stoichiometries.

dominated by near~t-neighbour interactions. in the middle temperature range so that the segrega-
expressed here by r&. Thus, only Ni atoms adjacent tion is two valued; the high value occurs if the tem-
to the grain boundary are used and. for Ni segrega- perature is attained by heating whereas a low one
tion, an equation similar to equation (3) may be occurs by cooling. These effects may occur in high
written: Ni content alloys with P segregation or in alloys
x?4i where the impurity and alloying additions are such
- = XcSiexp(AGki RT). (4) ,
1 - xxi as to make zyI htgh.
If Ni levels higher than one monolayer are permitted
2.4 The kinetic development of the segregation
there is. however, no marked change in behaviour
in any of the calculations that follow although the Analysis of the kinetic development of the equilib-
precise values of the various functions do alter. rium segregation level of P shown in Fig. 5 gives the
Equations (3) and (4). with the values from Table Tt plot of segregation directly. McLean, in 1957 [24],
2, may be readily sot\cd to give the level of segrega- derived, by diffusional analysis, the kinetics of grain
tion, at equilib~um. of the P in Carr et al. [t6] steel, boundary segregation expressed by:
as shown in Fig. 5. The segregation level of P rises
very rapidly as the temperature falls at around 500°C.
Also shown on figure 5 by the dotted lines are the
range of segregations that would be predicted for
adsorption in binary systems adjusted to the same
segregation at 500°C; (a). the segregation level for a
higher P content alloy using AG, instead of AG; and
which is appropriate to segregation in pure iron and, I \ -‘, ‘\ I
(b), the segregation level of the low P content ahoy
with a free energy of segregation of 13.2 kcal mol-‘.
These curves illustrate that the sharp change at 5OO’C
cannot be obtained for segregation in binary systems
even if AGr. can be arbitra~~y chosen to predict a
suitably high segregation level. The behaviour of both
of the dotted curves, although giving the same segre-
gation level at 500°C as the full curve, show only
a slow rise in segregation level as the temperature
falls through 5OO’C and is characteristic of segrega-
tion in binary systems [25].
Figure 5 shows the effect of the lowish coupling
constant, &.. As x’ increases or as the Ni or P bulk Fig. 5. The calculated segregation level of P as a function
levels rise the segregation levels increase and the tran- of temperature according to equations (3), (4) and Table
sition behaviour sharpens until, with x’ above about 2 is shown by the full line. The dotted curves represent
the calculations for the binary Fe-P system (a)
8 kcal mol-‘. there appears a discontinuous increase AG, = 10.5 kcal mol-’ with phosphorous content O.W&
in segregation level as the tem~rature falls. With and (b) AG, = 13.2 kcal moi-’ with phosphorous content
higher values of 2’ the Curve becomes three valued 0.01570 by mass.
352 SEAH: GRAIN BOUNDARY SEGREGATION

where Xb(Ol,X,(r) and X,(-c) are the grain boundary difficuft tu compare Fig. 6 with Fig. 1 or 1 in detail
segregation levels at the start of an isothermal heat- as the sharpness of the noses in the latter are a little
ing, time E and at infinite time (i.e. the equilibrium subjective. Instead horizontal and vertical sections of
level), respectively. D is the volume diffusivity of the the fz plots will be compared as this permits a direct
solute, I is the ratio of the concentration of the solute use of the original measurements rather than the in-
in the grain boundary to that in the grain interior terpolated plot.
and d is the grain boundary width occupied by the The Carr er al. Cl61 results for the DBTT shift from
segregant (the value of d does not matter since the the unembrittIed state, as a function of time for isoth-
product rd is the parameter measured and is equal ermal treatments at the critical temperature of 5oO’C
to the ratio of the atoms of segrgant per unit area going through the nose, and also at 475% where
of boundary to the atoms per unit volume of the there is also considerable data. may be accurately
bulk). The vaIue of D for P in pure iron is given retrieved from Fig. 2. The correlations of these results
by Grutin and Mural’ [44]: and the calculated data from Fig. 6 are shown in
Fig. 7. The calculated se~egation levels for the two
D = 1.58exp(-52300/RT)cm2s-‘. (6) temperatures are plotted with different abscissae so
Also, Gruzin and Mural’ [44] show that Mn increases that the equilibrium segregations occur at full scale
the diffusivity of P in iron but that MO and Cr retard in each case. Plotted in this manner the two curves
it. If the effects of these alloy elements are allowed have exactly the same shape but are merely shifted
for, pro rata, in SAE 3140, the calculated dflusivity relative to one another on the time axis. This helps
returns approximately to that expressed by equation to emphasise that the development of the segregation,
(6). and hence the emb~ttlement, in kinetic terms follows
Equations (5) and (6) and Fig. 5 may now be com- this universal curve independent of segregant, matrix
bined to generate a Tt plot of P segregation in SAE or temperature. Changes in segregant, matrix and
3140 analogous to the embrittlement diagram of Fig. temperature, only alter the position of the curve on
2. Carr er al. [16] started with material tempered for the time axis and the scale of the abscissa Superim-
I hr at 675C to provide the unembrittIed condition posed on the curves are the me~urements of Cat-r
with the correct strength level. From Fig. 5 the equi- et al. [16]. To get the best fit of the experimental
librium segregation level at 675°C is 0.08 monolayers results to the theoretical curves a shift of 275°C per
of P but from equation (5), after 1 hr, only 80% of monolayer of segregated P has been assumed in
this level will be attained. Thus, the starting condi- establishing the abscissae. This shift is very similar
tion, with a transition temperature of -83”C, has a to the value for the 3CriMo rotor steel discussed in
grain boundary segregation level of 0.06 monolayers Section 2.1. The excellent correIation of curve shape
of P. Some small segregation may occur during the and the relative magnitudes of the results at both 475
austenitizing treatment so that the starting level may and 500°C shows that McLean’s [24] d~~iona~
be over 0.06 monolayers of P but this small increase analysis of the kinetics of segregation is valid and
wit1 be offset here by the local depletion of the bulk that any errors in the predicted P segregation level
level around the grain boundary due to the shortness or in the use of the pure iron P diffusion coefficient
of the 675°C treatment. are very small.
In the above manner the ‘ft plot for the change To demonstrate the accuracy of the prediction as
in P segregation, following the temper for 1 hr at a function of temperature, vertical sections of Figs.
675”C, was calculated directly and is shown in Fig. 2 and 6 at 200 hr are shown in Fig. 8. Of course,
6. It is interesting to note that there is suggestion a shift of 275-C per monolayer of P has now to be
of a double nose in the segregation which is similar assumed. The correlation is excellent again. especially
to but weaker than that corrected out of the original in view of the fact that at the higher temperatures
plot of Jaffe and Buffum [14] shown in Fig. I. It is some microstructural changes such as grain
)

1 10 no 1oWJ 1ocwJo
Embrittting time. 2,”

Fig. 6. The time-temperature diagram for segregation of P in the Carr er al. [I63 SAE 3t40 jtd
derived from Fig. 5 and equations (5) and (6). The numbers against the curves denote the P segregation
level occurring as a result of the embrittkg treatment, i.e. in excess of the 0.06 monolayers segregated
during the tempering treatment.
 SEAH: GRAIlV BOUNDARY SEGREG.ATION 353

Fig. 7. The predicted extra segregation of P (abscissa scales marked on the inner edge of the frame)
during embrittiement at NO’C and 475C as a function of time with the experimental embrittlement
data of Carr et al. [16].

growth [17] must still have been occurring causing given realistic values assessed from separate experi-
an additional increase in DBTT. The bluntin: of the mental data.
tip of the nose is compatible with there being a nar-
row spectrum of grain boundary adsorption sites or 3. FURTHER PREDICTIONS OF THE DATA
a spectrum of gain boundaries with the parameters
of Table 2 exhibiting distributions with half widths Having obtained an excellent correlation with Carr,
of 30, of their mean values. Goldman, Jaffe and Buffum’s [16] work there are a
Thus, the adsorption theory proposed by Gutt- number of predictions that can be made using the
mann [73] and modified for the individual behaviours same data. It is possible to extend the time and tem-
of P and Ni together with the kinetic analysis of perature DBTT data to times as long as 30 yr or to
hIcLean [24] appears to explain, in detail, the time operating temperatures below 37X. to evaluate the
and temperature dependence of the classic temper shift in temperature and time of the position of the
brittleness measurements of Carr et al. [16] without nose as the impurity level is altered or to evaluate
recourse to further data. The only parameters to be its position in different alloys.
determined are AGp, AGNir x,kip and the DBTT shift The first predictions are evident in Fig. 6 already
per monolayer of P. Here all of these parameters are where the time scale exceeds 30 yr and where extrapo-

Trons~t~on temperaturc:C
_ _ _ _ _ _, 0 10 20 30 LO
700. 4, I I I I / I

I I
I
1 I
02 03 OL 05 0.6
~onobycrs of phosphorus

Fig. 8. The predicted extra segregation of P as a result of 200 hr embrittlement as a function of

isothermal holding temperature (full line) with the experimental embrittlement data of Carr et al. [16]
(dotted line).
354 SEAH: GRAIN BOUNDARY SEGRECiATION

larion to low temperatures is easy. It- is interesting lations for XISI 33-W show that the equivalent range
to consider that a rotor made from SAE 3140 of this will be 525-6OO’C. whereas for SNiCrMoV they will
purity level. even if manufactured with sufficiently fast be even higher. These predictions are confirmed by
cooling to reduce segregation will, if operated at the experiments of Rosenstein and Asche [45]. It
350°C. exhibit 0.4monolayers of P at the grain should be noted here that the Tt diagram predictions
boundary after the short life of 10 yr. Thus. the segre- are all for segregation and may only be converted
gation contribution to brittle failure and stress cor- to DBTT shifts for instances where the brittle fracture
rosion susceptibility can be predicted for very long mode is intergranular. The actual transition tempera-
operating lives reasonably accurately and, in this case, tures and the shift per monolayer of segregant depend
would be quite detrimental. on the alloy itself and its microstructure.
As the P impurity level changes the rapid rise in As the P purity level improves, the temper brittle
equilibrium segregation, which always occurs at nose falls in temperature and also the point A, an
around 0.5 monolayers, will change in temperature. indicator of the position of the nose shown in Fig.
At temperatures below that required to segregate 0.5 6, shifts to longer times. A is in fact at the temperature
monolayers the behaviour will be much as shown for at which the equilibrium segregation is 0.5 mono-
the lower temperatures in Fig. 2 except that the layers and at the time to reach half this level. The
DBTT shift denoted by each line will merely be times for point A are shown by the numbers against
adjusted pro ram for the change in bulk P level, i.e. the curve for SAE 3140 steel in Fig. 9. Thus, for in-
if the bulk level of P is halved, the values of the stance, even with the relatively low P content of
DBTT shifts are halved. Thus, because of the slow 0.008%, operation for one month at 468’C, or with
diffusion kinetics, in commercial rotor manufacture a P content of O.OOSo/, one year at 440X. the segrega-
where a moderately rapid cooling heat treatment tion would increase by 0.25 monolayers producing
occurs, the final DBTT of the rotor will be approxi- a DBTT of - 14°C. The full effect of improved purity
mately proportional to the bulk impurity content. At on the diagram is illustrated in figure 10. Curve A
the temperature for 0.5 monolayers of segregation a represents the 0.3 monolayers excess segregation Tt
rapid change in behaviour occurs and the upper edge contour for SAE 3140 steel, from Fig. 6, and here
of the nose is encountered. Thus, a full picture of is equivalent to a DBTT of OC. If the bulk P content
the Tt behaviour can be generated simply by a know- is halved to 0.0075:,/, the plateau region of the nose
ledge of the temperature dependence of the 0.5 mono- drops as discussed above and the lower, kinetically
layer segregation level as the bulk P level changes. limited part takes some four times as long to reach
This latter function can be easily evaluated from any given level. Thus, although at, say, 425°C the
equations (3~ and (4) and the data of Table 2 and equilibrium segregation level is relatively unchanged
is shown in Fig. 9 for the Ni-P interaction in SAE (near one monolayer) in a practical situation of
3140 and AISI 3340 and for P segregation in pure temper brittleness the material spends a given limited
iron. It is clear that commercial SAE 3140 with P time in any temperature range and it is the kinetics
levels in the range 0.008 to 0.025% will exhibit a which determines the embrittlement.
temper brittle nose with its upper edge always in the Thus, on improving the purity the nose shape of
narrow temperature range 475-530°C. Similar calcu- the Tt plot remains but is shifted. as shown, from

Fig. 9. The predicted temperature of the upper edge of the temper brittle nose as a function of P
impurity content for pure iron, SAE 3140, and AISI 3340 steels.
 SEAH: GRAIN BOUWDXRY SEGREGATIOY 3%

curve A to 3. Curve C shows how. in the halved JOO’C. other conditions being equivalent. There is no
impurity content material the halved segregation con- data for the effect of alloying elements on the diffusi-
tour remains much like curve A. It is this similarity vity of Sn in steel and so it will be assumed that
between curves A and C that produces, in slow cooled the pure iron data is valid. The interaction coefficients
or commercial heat treatments a degree of embrittle- of Sn are similar to those of P [13] and so. for in-
ment of rotor steels that is linearly or near linearly stance, an alloy with 0.029 wt.‘:; Sn (same atomic con-
dependent on the embrittling impurity content [46]. centration as 0.0075 wt.‘,;‘P, much higher than nor-
This linearity is not, of course. relevant to long term mal industrial practice) will exhibit a 0.3 monolayer
isothermal heat treatments and analysis of these cases contour lying in the series described by curves B and
must be made as described in Sections 2.3 and 2.4. D in Fig. 10 but with 2-3 times longer time. If r;,,
The effects of changes in the alloy composition are is high, the curve will be nearer D, if low. nearer
also shown in Fig. 10. If the Ni content of the alloy B. Measurements on a 5NiCrMoV rotor steel by the
with 0.0075% is raised to 3.5”;. as in AISI 3340, the author indicate that. in fact &,, is only a little higher
low temperature region remains unaltered but the than I&.
plateau is raised some 65-C so that the final nose It is clear from the genera1 appearance of Fig. 6
position is around 5YC. for 0.3 monolayers segre- that detailed equilibrium adsorption curves are not
gated, as shown by curve D. There are few sets of required to get a fairly clear picture of the Tr segrega-
measurements to confirm this behaviour but Rosen- tion diagram. Calculation of the temperature for i
stein and Asche[45] present results for a 30,; Ni a monolayer equilibrium segregation gives the pos-
1.650,,;,
Cr 0.4% MO steel with O.O@$~P which show ition of the top of the nose and estimation of the
the nose to be within lO’C of 53O’C, in agreement effects of kinetics at lower temperatures, according
with the above prediction. to the very simple Crank [48] equation:
Changes in the alloy element can cause changes
in both the P diffusivity and the interaction coefficient Jd’ p
z&. If the P di~usisity is increased or if oh is larger
X*(d - hm = x, -j- , -,
a V:: (7)
than skip, the nose in the Tr diagram shifts to higher
temperatures and higher P segregation levels will be where 0’ is the atomic volume of the matrix atom,
observed after industrial continuously cooled treat- will give accurate values for the times of segregation
ments. Similarly, changes in impurity type change the up to 0.2 monolayers. Of course, equations (.5) and
same two terms. In most UK steels only P and Sn [?I (7) are identical for low levels of segregation but use
residuals are at sufficiently high levels to cause of the latter removes the need to know the enrichment
embrittlement and so only Sn need be considered in ratio, X.
addition to P. The diffusivity of Sn in u-Fe, in the A final comment should be made about the effect
temperature range below 76o”C, is [473 given by of changes in the interaction coefficient, &. Xt low
D = 5.4exp (-555OO/R7). This gives a diEusivity values of &i the segregation behaviour is much like
about f that for P at 55O’C and -5 that at WC. that of a binary system with the segregation weakly
Thus the kinetically limited parts of the Tt plots, for temperature dependent. As c&i increases a sharper
Sn to build up any given segregation, take twice as and sharper increase in segregation occurs at a certain
long as for P at 500°C and three times as long at temperature. as the temperature falls. and. esentually,

____3X0,
DCAfSl
______ 75ppm.
______________“_________
03monl
____ _-

AISAE 3~0,lSoppm. 0~3mrn I

EWE3160, mpm. OJWSI

Time. hours

Fig. IO. The constant segregation 72 contours for Sc\E 3140 Steel: A: 190ppm P impurity contmt.
0.3 monolayers segregated; B: 75 ppm P impurity content, 0.3 monolayers segregated; C: 75 ppm
P impurity content. 0.15 monoiayers segregated; and. for AISI 3340 steel: D: 75ppm P impuriv
content, 0.3 monolayers segregated.
356 SEAH: GRAIN BOUNDARY SEGREGATION

the increase becomes discontinuous. As & is mcreas- 5. L. G. Emmer, C. D. Clauser and J. R. Low, w R
ing the alloy-impurity interaction becomes stronger, C. Bulletin 183, I (1973).
6. H. R. Tiplcr. fnr. Conj. on rhc Props of Hear Resisran)
the solubility of the impurity falls and the equilibrium
Steels. Dusseldorf, May (1972).
segregation level rises. However. if &, increases so 7. H. R. Tipler and D. McLean, .bfar. Sci. J. 1, 103 (1970).
that the solubility limit falls below the bulk impurity 8. P. R. V. Evans, N. B. Owen and B. E. Hopkins, Eng
concentration the equilibrium segregation will remain Fracr. .Vech. 3. 463 (1971).
9. P. E. Irving. M. P. Seah and A. Kurzfeld, Proc. 2na
high but, because of the declining free impurity level
Inr. Conf Mech. Behaviour of .Clarerials Boston, ASM
the kinetics of the segregation fall and in practical
563 (1976).
slow cooled heat treatments the final level of segrega- 10. R. P. Harrison and G. M. Spink, CEGB Rept. . RD/L,N
tion will be proportionately lower. At very high 174172 (1972).
values of & the solubility is so minute that no segre- 11. R. M. Latanision and H. Onoerhauser. Mer. Trans.
I.

gation is observed in practice. An example of this 5, 483 (1974).

12. K. Yoshino and C. J. McMahon, Mer. Trans. 5, 363
case is the MnS reaction in which z&s is very high (1974).
and so S segregation in temper brittleness can be 13. A. Joshi and D. F. Stein, Corrosion 28, 321 (1972).
eliminated by Mn treatment [X]. 14. L. D. Jaffe and D. C. Buffurn, Trans. .-IS&f 42, 604
(1950).
15. D. C. Buffum and L. D. Jaffe. Trans. ._(I.VfE 191, 540
4. CONCLUSIONS (1951).
16. F. L. Carr, M. Goldman, L. D. JatTe and D. C. Buffum
Use of a modified form of Guttman’s [23] theory Trans. AIME 197, 998 (1953).
of equilibrium gram boundary segregation of impuri- 17. B. C. Woodfine, J. Iron Sreel Insr. 173, 229 (1953).
18. J. R. Low Fracrrrre of Engineering .\larerials p. 127.
ties in alloy systems together with McLean’s analy-
(ASM Metals Park, OH) (1964).
sis [24] of the kinetics of segregation allows the gener- 19. C. J. Mchiahon, ASTM STP 407, 127 (1968).
ation of Tc segregation diagrams for low alloy steels. 20. A. E. Powers, J. lron Steel Inst. 186. 323 (1957).
Values for the free energies of segregation of P and 21. G. C. Gould, ASTivf STP 407, 90 (1965~.
Ni in iron and the relative interaction coefficient, xfNip, 22. D. McLean and L. Northcott, J. Iron Sreel Inst. 158,
169 (1948).
are all estimated from separate experimental results 23. M. Guttmann, Surface Sci. 53, 213 (1975).
and incorporated into the above theory to deduce 24. D. McLean. Grain Boundaries in &ferals. Oxford Uni.
the segregation 72 diagram for the SAE 3140 steel versity Press, London (1957).
used by Carr et a!. [16] in their evaluation of the 25. M. P. Seah and E. D. Hondros. Proc. R. Sot. London
embrittlement Tt diagram. If a reasonable value is (A) 335, 191 (1973).
26. E. D. Hondros and M. P. Seah, Inr. Mer. Rec. To
used for the DBTT shift per monolayer of segregated be published.
P the two TC plots correlate very closely over the 27. J. R. Low. D. F. Stein, A. M. Turkalo and R. P.
full range of times and temperatures. Thus the normal Laforce, TIMS-AlltlE 242, 14 (1968).
behaviour of temper brittleness, as exampled by the 28. F. L. Carr, J. Nunes and F. R. Larson. ASTM STP
407, 203 (1968).
above classic measurements, may be fully explained 29. D. Kalderon, Proc. Insr. Me&. Eng. 184, 341 (1972).
by the use of an equilibrium segregation theory with- 30. H. Ohtami, H. C. Feng, C. J. McMahon and R. A.
out reference to the effects of grain boundary car- Mulford, &fet. Trans. (A) 7, 87 (1976).
bides. 31. J. R. Relhck and C. J. McMahon, &fer. Truns. 5, 2439
(1974).
Predictions extended to other materials shows how
32. M. Henderson Brown. unpublished work (1976).
the temper brittle nose in the ?? diagram shifts to 33. R. Viswanathan and T. P. Sherlock. Met. Trans. 3.
lower temperatures and higher times as the purity 459 (1972).
level increases and that the exact reverse occurs as 34. R. Viswanathan, Met. Trans. 2. 809 (1971).
the nickel or alloy content increases. Both predictions 35. H. L. Marcus, L. H. Hackett and P. W. Palmberg.
ASTM STP 499. 90 (1972).
are supported by published measurements and 36. P. V. Ramasubramanian and D. F. Stem. Met. Trans.
demonstrate the power of this approach in the quanti- 4, 1735 (1973).
tative assessment,of segregation and embrittlement in 37. P. V. Ramasubramanian, Ph.D. Thesis University 01
practical situations. Minnesota (1971).
38. T. Inoue. Proc. 4th Bolton Landing Conf Grain Boun-
Acknowledgrmenrs-The author would like to thank Dr. darirs in Eng. Mar. p. 553 (1974).
M. Guttmann for discussions of his theory, Dr. R. A. Hunt 39. 0. Kubaschewski, E. L. Evans and C. B. Alcock
and Mrs. P. Cooper for the computation of the detailed Mernlluraical Thermochemisnv 4th Edn. Pergamon
shape of the segregation Tr diagram, Drs. E. D. Hondros Press, Oiford (1967). .
and C. Lea for their comments and Drs. F. R. Larson 40. M. P. Seah and C. Lea, Phil. .\lug. 31. 627 (1975).
and F. L. Carr for the supply of the SAE 3140 material 41. M. Guttmann. &feral Sci. 10, 337 (19761.
used in the appendix below. 42. H. Kaneko, T. Nishizawa K. Tamaki and A. Tanifuji
J. Japan Insr. &fer. 29, 166 (1965).
43. A. Joshi and D. F. Stein, ASTM STP 499, 59 (1972).
REFERENCES 44. P. L. Gruzin and V. V. Mural’, Fiz. MeraL Metakwed
17, (3). 62 (1964).
1. M. P. Seah, Surface Sci. 53, 168 (1975). 45. A. H. Rosenstein and W. H. Asche, tiT_Vf STP 407.
2. E. D. Hondros, M. P. Seah and C. Lea, .Cier& and 46 (1968).
Marerials January, p. 26 (1976). 46. G. C. Gould. ASTM STP 407. 59 (1968).
3. M. P. Seah, Proc. R. Sot. London (A) 349, 535 (1976). 47. D. Trehaux, D. Marchive, J. Delagiange and P. Guiral.
4. R. Bruscato, Wd J. 49, 148s (1970). denq, Cr. Acad. Sci. (C) 274, 1260 (1972).
 SExH: GR.4IX BOUNDARY SEGREGATIOS 357

4s. J. Crank. 7Yrhr.~f~rhrmancs 01 Dt$sion Ind Edit. and so was heat-treated as described in the oilgina work
Oxford Cniversity Press, London (19753. by Jaffe and Buflum [14] for 1 hr at 9WC, water
49. P. IV. Palmberg and H. L. Sfarcus. Trans. .X41 62. quenched. and again for I hr at 675-C and water quenched
1016 (1969). to provide the unembrittled state. The piece was split into
50. H. L. Marcus and P. W. Palmberg. TMS-.-U.VE 215. two parts which were heated for 200 hr at 500 and -175°C
1664 (1969). respectively to coincide with the conditions for Figs. 7 and
51. R. P. Laforce. D. F. Stein and A. Joshi. Trans. AS.cl S. The samples were then studied by Auger electron spec-
62. 779 (1963,. troscopy to check the actual levels of segregation at the
5’. N. C. hfacDonald. H. L. Marcus and P. W. Palmberg. grain boundary. Only P was significant and the measure-
Proc. 3rd Ann. SEJI Symp. Chicago (19701. ments showed P levels in good agreement *-vi&Figs. 6-Y
53. M. Guttmann. P. R. Krahe. F. Abel. G. Amsel. M. but Sn, Sb. As and Cu segregation levels below the instru-
Bruneaux and C. Cohen. Nrr. Trans. 5. I67 (1974). ment detection levels of 0.01. 0.01, 0.07 and 0.03 mono-
31. C. L. Smith and J. R. Low. Mrr. Trans. 5. 179 11974). layers respectively. This confirms the conclusion of Section
2.1 that P is most likely to be the main embattling agent
and also the numerical conclusions described in the rest
of the paper. Section 2.1 has been left unahersd with its
APPESDIS arguments to show the likelihood of P segregation in this
steel as the arguments are general and P is the most com-
After the above study was completed a very small piece monly. observed grain boundary embrittler in ferrous sys-
of the ori_ninal stock of the SAE 3t40 material used by tems [Z]. This scrv2s to emphasise the generality of the
Carr er ~1. [I61 was obtained from the Watertown Arsenal whole analysis in relation to temper brittleness in commer-
Laboratory. Bulk analysis of the material showed that the cial practice. This latter point is not trivial a.s a consider-
levels of embrittling impurities other than P were very low, able amount of work, especially in the LX-I. has been to
Sn and Sb were below 10 ppm and As was only 80 ppm. evaluate and characterise the effect of Sb and not P in
The material was in the original “as received” condition low alloy steels [27.30,31,43.39-541.