Professional Documents
Culture Documents
Grain boundary segregation of impurities like P, S, Sb, and others is the origin of temper
embrittlement of low-alloy steels. Till now it has been assumed that the segregation is determined
by an equilibrium segregation mechanism, but some open questions cannot be satisfactorily
explainedby the equilibrium segregation mechanism. In the present work, a combined equilibrium
and non-equilibrium segregation mechanism of temper embrittlement was established and some
open questions on temper embrittlement were addressed by means of the model. The model was
applied to phosphorus segregation in an Fe-0.3% (2-3.5% Ni-1.7% CrC0.06% P steel.
*Present address: Institute of Polymer Technology and Materials Engineering, Loughborough University of Technology, Loughborough,
Leicestershire LEll 3TU, UK.
62
as a function of the holding time t at temperature Tj 2.2. A method of calculating the segregation
when the segregation process of the complexes to level of impurities in the tempered sample
grain boundary from the grain centres is dominant. Equilibrium grain boundary segregation occurs
Using Equation 5, we may predict the non-equilib- mainly during tempering of the sample and non-equi-
rium segregation level of impurities in the sample librium segregation occurs mainly during quenching
quenched from a higher temperature and then tem- and tempering. Therefore, it is of considerable import-
pered at the tempering temperature. ance to calculate the segregation levels during quench-
As mentioned above, a critical time tr exists at some ing and tempering. This calculation has to be divided
temperature. At some temperature when the critical into two steps because the matrix during quenching is
time is longer than the effective time of impurity diffu- austenite and the matrix during tempering is ferrite.
sion corresponding to the cooling process, re, the The first step of the calculation is calculation of the
process in which the segregation is dominant occurs segregation level during quenching and the second
alone; whereas when tc is shorter than te, the process step is of that during tempering.
in which desegregation is dominant also appears. The
critical time tc is given by I-9, 13] 2.2. 1. Calculation of non-equilibrium
segregation level during quenching
B 2 ln(Dc/Di) As discussed elsewhere [9], any continuously cooling
tc - (6)
4~(Dr - Di) curve for a sample can be replaced by a corresponding
stepped curve, each step of which is formed by hori-
where 6 is a numerical factor and B is the grain size. zontal and vertical segments so as to calculate an
When the effective time is longer than the critical effective time te at some temperature for this cooling
time at temperature Tj, the process in which deseg- process. The effective time formula of a stepped curve
regation is dominant will take place. The segregation consisting of n steps at temperature T is given by [-9]
level of impurities at grain boundaries in this case,
Cb(t), is given by [-10]
te = i=1--ti exp
( EA(--T~-
Ti)~
k TTI J (8)
Cb(t) = Cg + l[Cb(t~) - Cg] [erf(-[4Di(dn_---/2r ) where EA is the activation energy for the diffusion of
complexes in the matrix and it is assumed here that EA
is given by the average value of the vacancy movement
tc)]l/z/lj~]
- e r f ( [ 4 D i ~ _"n/2 (7) energy Em and the impurity atom diffusion activation
energy; tl and Ti are the isothermal holding time and
temperature at the ith step of the stepped curve, re-
where t is the isothermal holding time at temperature
spectively. When the chosen steps are small enough,
Tj, tc = tc(Tj) and Di is the diffusion coefficient of
the effective time corresponding to the stepped curve
impurity atoms in the matrix. Evidently, Equation 7 is
at some temperature will accurately enough be equal to
only concerned with desegregation. Thus the condi-
that corresponding to the continuously cooling curve at
tion t > tc is necessary for Equation 7.
the same temperature. It may be expected that the
According to previous work [-9, 13, 14], Equation
effect of diffusion of impurities (or complexes) during
6 is only suitable for a high temperature range in
the continuous cooling is the same as that of impu-
which the maximum equilibrium segregation level is
rities (or complexes) during isothermal holding for the
very small. At lower temperatures, since the maximum
effective time during cooling at the same temperature.
equilibrium segregation level, Coo is larger and the
This means that it is possible that calculation of the
process in which desegregation is dominant may be
segregation levels of complexes during quenching may
able to take place when the segregation concentration
be changed into calculating that during isothermal
is larger than Coo. Consequently during tempering of
holding at some temperature (chosen as 600 ~ in this
steels, the process in which desegregation is dominant
work) by means of the effective time concept. In this
cannot occur before the segregation level to grain
work T in Equation 4 was chosen as 600 ~ in calcu-
boundaries reaches the maximum equilibrium segre-
lation of the segregation level during quenching. This
gation one, Coo. In the present work, the grain bound-
is because it is assumed here that very little diffusion
ary segregation level of P before reaching C~ was only
calculated, which is evidently consistent with the real will occur below 600 ~ due to a larger quenching rate.
For any sample quenched, we can gain the effective
case. Furthermore, since non-equilibrium segregation
time corresponding to its quenching at 600~ as
is a process of kinetics and occurs mainly during
above. The segregation levels in the quenched sample
quenching and tempering of the sample, while equilib-
can be obtained by substituting the effective time into
rium segregation is a process of thermodynamics and
Equation 5. It is noted that C~(T,) in Equation 5 is
occurs mainly during tempering, the equilibrium seg-
about equal to Cg because the quenching temperature
regation level during tempering and the non-equilib-
rium segregation one during quenching + tempering is higher.
were calculated in this work. It should be noted that in
the present work it is assumed that the diffusion of 2.2.2. Calculation of segregation level after
impurity atoms to the grain boundary does not affect tempering
that of impurity atom-vacancy complexes to the grain After the non-equilibrium segregation level during
boundary and vice versa. quenching is gained as above, the equilibrium and
63
non-equilibrium segregation levels after tempering percentage. The non-equilibrium grain boundary seg-
may be obtained by substituting the tempering time regation width d = 100 nm was chosen. Dc was given
into Equations 2 and 5, respectively. In this calcu- by the product of the frequency factor for vacancy
lation, Dc and Di are the diffusion coefficients of diffusion and the exponential term containing the
complexes and impurities in the matrix (ferrite), average value of vacancy movement energy and solute
respectively. The grain boundary concentrations of diffusion activation energy in the matrix.
segregant at tempering time t = 0 at the tempering The results of calculating P segregation during
temperature, Co~(Ti) in Equation 2 and C~(Ti) in quenching + tempering are shown in Figs 1 and
Equation 5, should be equal to the non-equilibrium 2 which indicate the following:
segregation level during quenching. Finally, the total
1. The segregation level during quenching is about
level of impurity segregation after tempering is equal
4 a t % for the sample quenched from 1300~ (see
to the sum of levels of the two types of segregation
minus C,. Fig. 1), which is in agreement with the results of Ogura
et al. [16].
2.3. Theoretical results 2. At the same quenching temperature, the process
Mulford et al. [15] have shown bond weakening by of segregation kinetics in the sample tempered at
phosphorus segregation at grain boundaries ina 3.5% 500 ~ is much faster than that in the sample tempered
Ni-1.7% Cr-0.06% P steel. Therefore let us apply at 450 ~ which is well known, but it should be noted
that at the tempering temperatures of 500 and 450 ~
now the above model to P in a 3.5% Ni-1.7% Cr steel.
The composition of the steel, taken from Ogura et al. the kinetic processes of segregation for this model are
[16], is shown in Table I. all much quicker than those for McLean's model (see
Nowadays, most of the researches with respect to Fig. 1).
temper embrittlement are done by the following heat- 3. At the same tempering temperature and time, the
treatment. The heat-treatment steps of the sample are segregation level in the sample quenched from 1300 ~
always quenching + toughening above 600 ~ and is larger than that in the sample quenched from
then embrittlement between 350 and 600 ~ which, 1050 ~ (see Fig. 2), which cannot be interpreted by
the equilibrium segregation mechanism.
generally speaking, is not consistent with actual heat
treatments and also not suitable for the above model.
Consequently in the present work, theoretical results 3. Discussion
can only be obtained from the above model and they A new interpretation of temper embrittlement results
are not compared with observed data. from the diffusion of impurities and impurity-vacancy
It is assumed that the heat-treatment steps of complexes to grain boundaries from grain centres.
a sample 10 mm x 5 mm x 5 mm in size are
25
(i) austenitizing at 1300 and 1050~ respectively O0~
and quenching at a rate of 250 ~ s-*; and
(ii) tempering at 500 and 450 ~ respectively, for up 20
to 100 h.
The data used in calculating P segregation level
r
during quenching and tempering are given in Table II. 0
The vacancy-phosphorus binding energy Eb was
chosen as Eu = 0.36 eV. The concentration of P in the O)
grain centres, Cg, was taken as the analysed concen- t/)
64
25
~ 1300~ with increasing tempering time, the segregation level
of P in the sample tempered at 450 ~ will gradually
increase and be larger than that in the sample tem-
o~ 2O pered at 500~ The above results have well been
v
known, but it should be noted thatat temperatures of
500 and 450 ~ the kinetic processes of segrega.tion for
E
the model are all far faster than those for McLean's
O
model. This is because 9/10 of the time necessary for
glO reaching the maximum equilibrium grain boundary
segregation concentration, t9/10 , is given by [7]
O3
5 t9/lo = 468~2d2/64Di (9)
where d is the width of the grain boundary which is
0 10 2'0 3'o 4'o so 60 about 1 nm, Di is the diffusion coefficient of impurity
Tempering time (h) atoms in the matrix and ~ is the enrichment ratio of
Figure 2 Segregation levels of phosphorus as a function of impurities. For the ~-Fe matrix and P in it, ~ is about
tempering time in samples quenched at 250~ -1 from 1300 and 224 at 500 ~ and 301 at 450 ~ and Di is given by
1050~ and tempered at 500 ~ (Co = 25 at%). Table II. Therefore, substituting for d, ~ and Di in
Equation 9 gives t9/~o = 1311 h at 500 ~ and 28108 h
at 450 ~
Theoretically, this model is reasonable. This is because From the above discussion, we may see that excess
the excess saturation degree of vacancies in a steel saturation vacancies in the quenched sample play an
quenched from a higher temperature is very large. In important role in the tempering process of the steel,
this case, a large number of impurity-vacancy com- whereas this was not considered in the equilibrium
plexes would appear because the vacancy and the segregation mechanism of temper embrittlement.
impurity atom possess a certain binding energy. When The theoretical results in Fig. 2 show that for the
the steel is tempered, a loss of vacancies along grain same tempering temperature and time, the segregating
boundaries, i.e. at vacancy sinks, will be exhibited in level of P in the sample quenched from 1300 ~ is
order to reach the equilibrium vacancy concentration larger than that in the sample quenched from 1050 ~
at the tempering temperature. Therefore a complex which cannot satisfactorily be explained by the equi-
concentration gradient will appear between the grain librium segregation model but can be explained by the
boundary and the grain centres. The gradient of com- model suggested in this work. This is because the
plex concentration drives the complexes to diffuse to higher the quenching temperature is, the larger is the
the grain boundary from regions remote from it. This excess saturation degree of vacancies in the quenched
diffusion of complexes results in excess impurities con- sample, and the larger is the non-equilibrium grain
centrating near the grain boundary and leads to non- boundary segregation during quenching and temper-
equilibrium grain boundary segregation. Of course, ing. The above results can be used to interpret why an
the above processes also occur during quenching. intercritical heat treatment in the ~-7 region after
Consequently during quenching and tempering of the austenitization but before tempering may be used to
sample, impurity-vacancy complexes diffuse to grain reduce temper embrittlement [4].
boundaries to cause non-equilibrium segregation at Moreover, Fig. 2 also shows that the quenching
the same time as impurities diffuse to grain boundaries temperature apparently affects the kinetic process of
to cause equilibrium segregation. However, since the tempering. When the sample is tempered at 500 ~ the
equilibrium segregation is a process of thermo- segregation level of P reaches approximately the max-
dynamics, the segregation level of the impurity during imum equilibrium segregation level at a tempering
quenching of the sample is very small and may be time of 25 h for a quenching temperature of 1300 ~
neglected. but at 60 h for a quenching temperature of 1050 ~
The theoretical results in Fig. 1 show that the segre- This is because the excess saturation degree of va-
gation level of P at tempering time t = 0, i.e. during cancies in the sample quenched from 1300 ~ is much
quenching, is about 4 at % which is generally consis- larger than that in the sample quenched from 1050 ~
tent with the observed result of approximately 5 at % Since the segregation of P at prior austenite grain
[16] and reaches the maximum equilibrium segrega- boundaries is the accumulation of equilibrium and
tion level of about 24 at % at the tempering time of non-equilibrium segregation in the two processes of
25 h for a sample quenched from 1300~ and tem- quenching and tempering, whereas the segregation at
pered at 500 ~ However, the segregation level of P at ferrite grain boundaries occurs only during tempering,
a tempering time of 90 h for the sample quenched from the segregation level of P at prior austenite grain
1300 ~ and tempered at 450~ only reaches approx- boundaries is much larger than that at ferrite grain
imately 21 a t % , which is much less than the max- boundaries. Consequently the intergranular fracture
imum equilibrium segregation level of about 33 at %; caused by impurity segregation occurs predominantly
at the same quenching temperature and tempering at prior austenite grain boundaries, which was re-
time, the segregation level of P in the sample tempered ported by Briant and Banerji [ 4 ] .
at 500~ is much larger than that in the sample The width of the concentrated layer along grain
tempered at 450 ~ However, it may be thought that boundaries, d,, was chosen as 100nm in the
65
calculations of the present work. This is because the 5. K . T . AUST, S. J. A R M I J O , E. F. K O C H and J. A. WEST-
non-equilibrium segregation is a kinetic process. B R O O K , Trans. Amer. Soc. Met. 60 (1967)360.
6. T . R . A N T H O N Y , Aeta Metall. 17 (1969) 603.
Therefore, the segregation extent isgenerally larger
7. D. McLEAN, "Grain Boundaries in Metals" (Oxford Univer-
and may be up to tens of millimicrometres [13, sity Press, Amen House, London, 1957) pp. 118, 131.
20-23]. 8. XU T I N G D O N G , J. Mater. Sei. Lett. 7 (1988) 241.
The coefficient of diffusion of complexes in the 9. SONG S H E N H U A , XU T I N G D O N G and YUAN ZHEXI,
matrix, De, is given by the product of the frequency Aeta Metall. 37 (1989) 3t9,
10. XU T I N G D O N G and SONG S H E N H U A , ibid. 37 (1989)
factor for vacancy diffusion and the exponential term
2499.
containing the average value of the vacancy move- 11. R . G . F A U L K N E R , Mater. Sei. Teeh. 1 (1985) 442.
ment energy and the impurity diffusion activation 12. M . A . V . C H A P M A N and R. G. F A U L K N E R , Aeta Metall.
energy in the matrix. This is reasonable because the 31 (1983) 677.
diffusion of complexes is related not only to the move- 13. R . G . F A U L K N E R , J. Mater. Sei. 16 (1981) 373.
14. XU T I N G D O N G , SONG S H E N H U A , SHI H U A Z H O N G ,
ment of vacancies but also to that of impurities.
W. G U S T and YUAN ZHEXI, Aeta Metall. Mater. 39 (1991)
3119.
15. R . A . M U L F O R D , C. J. M c M A H O N Jr., D. P. P O P E and
4. S u m m a r y H. C. F E N G , Metall. Trans. 7A (1976) 1183.
A new mechanism of temper embrittlement, a com- 16. T. O G U R A , C. J. M c M A H O N Jr, H. C. F E N G and
V. VITEK, Aeta Metall. 26 (1978) 1317.
bined equilibrium and non-equilibrium grain bound-
17. G. SEIBEL, Mere. Sei. Rev. Met. 61 (1964) 413.
ary segregation mechanism, has been developed. 18. F . S . B U F F I N G T O N , K. H I R A N O and M. C O H E N , Acta
Theoretically, some open questions on temper embrit- Metall. 9 (1961) 434.
tlement may satisfactorily be explained. The cal- 19. T. O G U R A , Trans. Jap. Inst. Met. 22(2) (1981) 109.
culation results obtained by the new model agree 20. T. M. W I L L I A M S , A. M. S T O N E H A M and D. R. HAR-
RIES, Metal Sei. 10 (1976) 14.
satisfactorily with the existing observed data [16]. 21. R . G . F A U L K N E R and T, C. H O P K I N S , X-ray Speetrom.
6 (1977) 73.
References 22. P. D O I G and P. E. J. F L E W I T T , Aeta Metall. 29 (1981) 1831.
23. R . G . F A U L K N E R , ibid. 35 (1987) 2905.
1. W. S T E V E N a n d K . B A L A J I V A , J. lronSteellnst. 193(1959)
141.
2. M . P . SEAH and C. LEA, Surf Sci. 53 (1975) 272.
3. M . P . SEAH, Acta Metall. 28 (1980) 955.
4. C . L . B R I A N T and S. K. BANERJI, Int. Metall. Rev. No. 4 Received 15 May 1992
(1978) 164. and accepted 6 July 1993