Journal of Nuclear Materials 98 (1981) 173-189 173

North-Holland Publishing Company

CREEP AND CREEP RUPTURE OF ERNiCr-3 WELD METAL *

R.L. KLUEH and J.F. KING

Metals and Ceramics Division. Oak Ridge National Laboratory, Oak Ridge, Tennessee 37830, USA

Received 9 October 1980

Creep and creep-rupture tests were made on AWS A5.14 Class ERNiCr-3 weld metal, commonly known as Inconel82.
Specimens from gas tungsten-arc welds were tested over the range 454-732°C. Tests at 454, 510, and 566°C displayed
different characteristics from those at 621,677, and 732°C. The lower temperature creep curves showed a rapid transient, a
long steady-state stage, and little tertiary creep. At elevated temperatures the transient stage was slower, the steady-state stage
was shorter, and the tertiary stage was longer. Creep-rupture curves and stress-minimum creep rate curves were much flatter
at lower temperatures. Certain tests at 454, 5 10, and 566°C exhibited an “instantaneous elongation” or “strain burst”
phenomenon, in which the creep curves contained strain jumps. At these same temperatures several specimens failed
prematurely. The difference in behavior in the two temperature regimes was attriiuted to short-range order.

1. Introduction 100, 1000, and 10 000-h rupture stresses at 540,650,

760,870, and 980°C but gives no actual experimental
The nickel-base alloy with the American Welding results. The only other known creep data are
Society designation AWS A5.14 Class ERNiCr-3 is a unpublished proprietary creep-rupture data of
commonly used filler metal for joining several nickel- Huntington Alloys [4], data to which we had access.
containing alloys by a variety of welding processes. The Huntington Alloys creep-rupture data were usu-
Its nominal composition is as follows: 67% Ni, 20% ally in the form of two or three data points - enough
Cr, 3%Mn, 2.5% Nb, and lesser amounts of several to establish a rupture curve and estimate stresses for
other elements. This alloy, also known as Inconel 82, rupture in 10000 h - at 538,649, 760, and 87l’C.
the trade name of Huntington Alloys, has found No data are available at the temperatures applicable
wide application as a filler metal for dissimilar-metal to the operation of the transition joint - near 5 10°C.
joining. Its widespread use for joints between austeni-
tic stainless steels and ferritic steels makes it a prime
candidate for use in liquid metal fast breeder reactor 2. Experimental procedure
plants, where the intermediate loop piping is to be
constructed of type 316 stainless steel and the steam Two types of welds were made. The first was made
generators of 2$Cr-1Mo steel [ 11. Previous service between two 13-mm-thick plates - one plate of
experience with such joints in fossil-fired plants has 2@-1Mo steel and the other of alloy 800H. A 90”-
caused sufficient concern that a program was initiated included-angle V-groove geometry was used, requiring
to improve the joint life expectancy. a root pass and seven fill passes. Since very few speci-
We previously published data on the tensile mens could be obtained from such a small weld, a
properties of ERNiCr-3 weld metal [2]. Few creep larger weld geometry was designed. Plates of 19-mm-
data are available in the literature: Payne [3] gives thick 2iCr-1Mo steel were welded with a 30’.
included-angle V-groove joint geometry with a 32-mm
* Research sponsored by the Division of Reactor Research root opening and a backing strip. To till this joint
and Technology, US Department of Energy, under Con- required 40 weld passes. In both cases the welding
tract W-7405eng-26 with the Union Carbide Corporation. was performed by the automatic gas tungsten-arc

0 022-3 115/8 1/0000-0000/$02.50 0 North-Holland

174 R.L. Klueh, J.F. King / Creep and creep rupture of E’RNiCt-3 weld metal

process with cold-wire filler additions. The nominal mechanical extensometers attached to the specimen
welding parameters were 150 A and 10 V with a grips, and the specimen elongation was read periodi-
travel speed of about 0.9 mm/s. Table 1 gives the cally on a dial gage.
chemical composition of a large and a small weld, as
well as the Americal Welding Society (AWS) and
American Society for Testing and Materials (ASTM) 3. Results
specification for ERNiCr-3. Because of base metal
The first tests were made at 510 and 566’C on
dilution, the small weld contained considerably more
specimens taken from the smaller weldments (seven-
iron than the large weld.
pass welds made on 13-mm-thick plates). Because
All welds were radiographed, and no defects were
only a few specimens could be obtained from the
detected. Because of several premature creep failures
small weldments, later tests were made on specimens
early in the program, all subsequent specimens were
taken from the large weldments (40-pass welds on
radiographed before testing. Only a few specimens
19-mm-thick plates). The effect of the weldment size
contained defects; none of the defects were large
on the tensile properties was previously investigated
enough to warrant rejecting the weld. However,
[2] ; the effects that this could have on the creep-
specimens that contained defects were not tested.
rupture properties will be discussed below. Several
The creep-rupture test specimens were a button-
specimens tested at 454, 510, and 566’C failed
head type with a 3.1%mmdiam. by 28.6~mm-long
prematurely and are not included in the graphical
gage section. The specimens tested were longitudinal
presentations of the data.
all-weld-metal specimens. Before testing, all speci-
The creep and rupture behavior were quite tem-
mens were postweld heat treated 1 h at 732’C.
perature-dependent. Two temperature regimes were
Creep-rupture tests were made at 454, 510, 566,
noted: The results at 454, 510, and 566°C differed
621, 677, and 732°C in air in constant-load lever-
from those at 621, 677, and 732°C. Logarithmic
arm creep frames. During testing the specimens were
plots of stress against rupture life (fig. 1) and stress
heated in a resistance tube furnace, and the tempera-
against minimum creep rate (fig. 2) indicated a
ture was monitored and controlled by three chro-
distinct difference in behavior for the two tempera-
mel-P versus alumel thermocouples attached to the
ture regimes. At the higher stresses and at 454, 510,
specimen gage section. Temperature was contolled to
and 566’C, the curves (figs. 1 and 2) were quite flat.
*l’C, the temperature along the gage length varying
At 510 and 566’C, there appeared to be changes in
less than *2”C. Elongations were determined with
slope (it broke downward) as the stress was lowered.
At these temperatures the curves for rupture life and
minimum creep rate both showed such breaks. At
Table 1 454”C, however, only the minimum creep rate curve
Chemical composition of ERNiCr-3 fiffler metal broke downward; that is, rupture-life data were not
Element Chemical composition (wt%)
sufficient to determine the break.
The rupture-life data at the high temperatures -
Small Large ASTM B.304 621, 677, and 732’C - had a different character. At
weldment weldment AWS A5.14 677 and 732’C a single straight line appeared to fit
the data. Although more data are required to confirm
Ni Balance Balance 67.0 min
cu 0.3 05 max the interpretation, at 621’C the curves appeared to
Mn 2.7 2.0 25-35 break downward. The minmum creep rate data
Fe 10 2.8 3.0 max showed this more clearly than the rupture-life data.
si 0.2 0.5 max However, the high-stress end of the 62l’C curves was
C 0.03 0.02 0.1 max
not as flat as the low-temperature curves.
S 0.015 max
Ti 0.3 0.41 0.75 max Creep rate data (fig. 2) are generally fit by a pow-
Nb 1 .a 1.8 2.0-3.0 er law equation of the type
Cr 18.0 19.5 18-22
P=a”, (1)
 R.L. Klueh, J.F. King / Creep and creep rupture of ERNicZ-3 weld metal 175

I 1 I I I1111 I I II lllll I 1 I I IllI] I I I Illll) I I I l--n-T’00

600 -

400 -

“566 .C
IOO- 0621 OC
.\ l 677 T
80-
l 732 T_,.
60. ’ ’ ” ““’ ’ ’ Lou”” ’ ’ ’ I”“’ ’ ’ ’ I”“’ ’ ’ ” ‘IlL
IO0 10’ 102 i03 !04 IO5
RUPTURE LIFE (hr)

Fig. 1. Creep-rupturecurvesfor ERNiCr-3(Inane1 82) weldmetal.

where u is the stress, C the minimum creep rate, and data curves all had n values between 24 and 28. At
A and n are constants. To characterize the data the high stresses very large n values were calculated.
better, we estimated n for the data in fig. 2 (table 2). Another indication of the difference in behavior at
It should be noted that these are but rough estimates the high and low temperatures was seen by comparing
based on the visual fit of the data. the creep curves at the various temperatures (fig. 3).
The estimated n values again reveal the difference At the low temperatures (fig. 3a) and especially at
in behavior at low and high temperatures. At 621, low stresses the curves were characterized by a high
677, and 732’C, the high-temperature regime, n is initial creep rate followed by a rapid transient during
between 8 and 10. For the data at 621°C, which were which the creep rate decreased to a “steady state”.
judged to have a break in the curve, this value applied Once the steady state was reached, it continued
at low stresses. The low-stress and low-temperature almost to failure. That is, these tests showed essen-

600

60 -3 -2
tcrs 10"
MIN:I%UM CREEP RA;F (%/hr)

Fig. 2. Relationshipbetween stress and minimum creep rate for ERNEr3 (Inconel82) weld metal.
176 R.L. Klueh. J.F. King / cleep and creep rupture of ERNicZ-3 weld metal

Table 2 3.0
I I I I i I
Estimated stress exponents for the creep of ERNiCr-3

Temperature Estimated Stress exponent, n

ec> transition
stress a) Hish Low
(MW stress stress

454 480 66 24
510 430 70 28
566 380 45 28
621 270 18 9.6
677 7.8
732 8.6
0.5 -
a) Transition between high-stress and low-stress regimes. a

I I I I I I
0
0 2 4 6 (~10~)
TIME (hr)

tially no tertiary creep. At the highest stresses the

transition from the primary to the secondary stage
was less abrupt.
At the two highest temperatures, 677 and 732’C,
the creep curves are characterized by very little
primary creep (fig. 3b). The test quickly reached the
steady&ate stage, which was also shorter than it was
at the lower temperatures. These curves have well-
defmed tertiary stages. Finally, the curve shapes at
621*C appeared to fall between those at the low tem-
peratures and those at 677 and 732’C. The transition
to the steady state occurred over a longer period of
time and with more strain than had occurred at 677
and 732°C. However, the curves at 62l’C were also
different from those at the lower temperatures in
that the initial creep rate was lower and the transition
to the steady-state creep rate was more gradual than
in the low-temperature tests.
The characteristic shapes of the creep curves are
further demonstrated by comparing the tertiary
creep behavior at the different temperatures. For
numerous alloys the time to the onset of tertiary
creep, tz, is related to the rupture life, t,, according
to
t2 = APR , (21 TIME (hr)

where A and ty are constants 15-71; <Yis very often

equal to unity, making the ratio ta/tn a constant. Fig. 3. Examples of characteristic creep-zurve shape in the
(a) low- and (b) high-temperature regimes. (a) Test at 379
As stated above, the tests at the [Link]
MPa at 510°C; test failed abruptly at 6087 h with a 5.3%
showed essentially no tertiary ’ creep. This was creep strain. (b) Test at 103 MPa at 732’C; only the early
reflected in the ratio tz/fR at 454, 510, and 566*C: portion of this curve is shown (a continuous curve was
the ratio was always greater than 0.92, and with a few obtained to failure at 634 h with 23% creep strain).
 R.L. Klueh, J.F. King / Cheep and creep rupture of ERhW-3 weld metal 177

exceptions was always greater than or equal to 0.97. In general the total elongations and reductions of
The exceptions occurred for short-time tests, in area were consistent at all temperatures (figs. 4 and
which a lack of data made it difficult to estimate 5). Except at 454’C, both decreased with decreasing
tz. In other words, the creep tests in the low-tempera- stress (increasing rupture life). At 454’C ductility
ture regime spent most of their lifetime in the pri- changed very little over the range of stresses tested.
mary and secondary stages. Interestingly, of the high-temperature tests (fig. 5)
At the two highest temperatures, 677 and 732’C, those at.621°C showed the least elongation.
the tertiary creep behavior differed considerably from As stated above, several specimens ruptured
that at the lower temperatures: here t&R was much prematurely at 454, 510, and 566’C. We also noted
smaller. While it approached 0.99 at the lower tem- another peculiarity at these temperatures: it can be
peratures, it approached 0.15 at 677 and 732OC;that described as an “instantaneous elongation”, a “strain
is, most of the creep strain occurred during the ter- burst”, or a “spontaneous deformation”. The
tiary stage. phenomenon exhibited itself in the form of a jump in
The results at 621°C fell between those at the the creep curves, examples of which are shown in
lower and higher temperatures but most resembled fig. 6. These “strain bursts” appeared at different
the results at the lower temperatures; for most tests times in the creep process: shortly after loading dur-
tJtR exceeded 0.9. ing primary creep and during secondary creep. Single

60
1 I1111111 I 11111111 I I I I11111 I IllI-

50
z

01 I I I1”111 I I I111111 I I1111111 I I IllllJJ

100 10’ IO2 IO’ IO’

RUPTURE LIFE (hr)

Fig. 4. Total elongation and reduction of area as functions of rupture life for ERNiCr-3 (Inconel82) weld metal at 454,510, and
566’C. Data points were connected to indicate the trends.
178 R.L. Klueh, J.F. King/Creep and creep rupture of ERNict-3 weld metal

2 5 402 2 5 40’ 2 5 404

RUPTURE LIFE Ihr)

Fig. 5. Total elongation and reduction in area as functions of rupture life for ERNiCr-3 (Inconel82) weld metal at 621,677, and
732°C. Data points wcze joined to indicate the trends.

and multiple bursts were observed. Fig. 6 shows a than 6%. When the strain burst occurred during the
creep curve that contained multiple bursts, some in steady-state stage, the creep rate after the burst was
the primary stage and some in the secondary stage. At usually little different than before it (fig. 6).
454OC, 73% of the tests displayed one or more No direct evidence could be found to link the
bursts, while 67% displayed bursts at SlO’C. Only strain burst phenomenon with the observations of
40% of the tests at 566’C displayed the phenomenon. premature failure. However, the fact that both
The tests without bursts at 454, 510, and 566’C phenomena occurred only in the low-temperature
occurred for the highest stress tests. No strain bursts tests could signify a relationship. It is as if a prema-
were seen at 621, 677, and 732’C. Strain burst mag- ture failure occurred during a strain burst larger than
nitudes ranged from strains of about 0.1% to greater the ductility of the alloy could absorb. The difficulty
 R. L. Klueh. J.F. King / Creep and creep rupture of ERhWr-3 weld metul 119

to conclude that such rupture lives were not prema-

ture failures.
Visual examination of the fractured specimens
revealed several fracture modes as well as two types
of anisotropic deformation. In many cases the sur-
faces appeared “fibrous” or “rumpled*’ with
“stretcher marks” parallel to the specimen axis. Some
specimeas also appeared “knobby”, as if one or more
points on the specimen gage section were in the pro-
cess of forming a neck. This “necking” was in addi-
tion to the point where the specimen necked to fai-
lure. Another indication of anisotropic behavior
was the observation that specimens often drew to an
2s elliptical cross section instead of to a circular one.
The tests at 454’C formed slight necks and in all
8. cases exhibited the anisotropic behavior described
above. Only at this temperature was there a pro-
nounced tendency to deform to an elliptical cross
3 section. Although at 510°C the “knobby” defor-
mation still occurred, actually very little neck forma-
2
tion took place at the fracture. in most cases the
fractures at 454 and 51O’C appeared to occur by
shearing, the fracture surface making about a 45’
t angle with the specimen axis. At the lowest stresses at
I SIO’C, however, the fracture surface made about Q
90’ angle with the specimen axis. The fracture sur-
Ok
0 B0.4 TIME (hr)
0.8 face contained “striations*’ in the direction of the
shear.
Fig. 6. Exampleof multiple“strain bursts” in ERNiCr-3 weld The observations at 566’C resembled those at
metal tested at 454°C and 489 MPa.
5 10°C. We observed unusual behavior only in the
[Link] test that failed prematurely. In this test
fracture occurred after 778.8 h, whereas a repeat test
in relating strain bursts to premature failures was failed after 6003.3 h. When the fracture surface was
evident when three essentially identical tests made at examined by stereomicroscopy, there appeared to be
454°C and 428 MPa were compared. Two of the tests a hole on the surface. Protrusions visible inside the
failed in less than 80 h, while the other test ruptured hole may have been dendrite tips, suggesting a defect
during a tempe~ture excursion at 2450 h (rupture formed during welding. Although the weld itself had
would be predicted from fig. 1 to occur in about been radiographed before testing, this specimen had
5000 h). One of the tests that failed prematurely not.
contained a burst of about 6% before failure. Both At 62 1“C all but one of the specimens had quite flat
the premature faiIures had loading strains greater than fractures with virtually no necking. The test at 379
22%, whereas the 2450-h test strained only 14% on MPa, the highest stress used at 62 1*C, formed a slight
loading. The latter test contained a strain burst early neck and exhibited more anisotropic deformation
in the test (during primary creep) of about 4% and than the other specimens did. Anisotropic deforma-
three short bursts of less than 0.5%. Tests that con- tion -both “knob” and “stretcher mark” forma-
tained at least one strain burst may well have had tion - decreased with decreasing stress and was
premature failures compared with those found in absent from the lowest stress tests. Tests at 677 and
tests without any bursts. Our results do not permit us 732’C showed the same behavior as at 62 l°C; that is,
180 R.L. Klueh, J.F. King / Creep and creep rupture of ERNict-3 weld metal

Fig. 7. Scatteredgrain boundary cracks on specimen tested at 396 MPa at 566’C. (a) Unetched. (b) Etched with 5 : 1, HCl
HN(D3.

only at the highest stresses were necks and anisotropic observed for most of the tests at the three lowest test
deformation observed. The fracture surfaces of tests temperatures.
at 621, 677, and 732’C were “striated” like those at Radiography was used to examine several unfrac-
the lower temperatures, but they all made a 90” angle tured specimens to determine whether strain bursts
with the specimen axis rather than the 45’ angle were accompanied by crack formation. Specimens
 R.L. Klueh, J.F: King / Oeep and creep rupture of ERhVCY-3weld metal 181

Fig. 8. Grain boundary cracks on specimen tested at 413 MPa at 5 10°C. Etched w-ith 5 : 1, HCl :HNOJ.

with strain bursts of up to 5% were examined. This

technique can detect cracks or voids of 0.08 to 0.13
mm, but we observed none.
Selected fractured specimens were examined
metallographically. At all temperatures cracks were
detected in grain boundaries back from the fracture
surface. For all stresses at 454°C and for the highest
stresses at 5 10 and 566°C only a few grain-boundary
cracks were detected, and they were widely scattered
on the specimen gage section (fig. 7). The number of
grain-boundary cracks increased with a decrease in
the creep stress at 510 and 566’C. For the lower
stress tests at 510 and 566°C we observed flat frac-
tures with essentially no necking and little deforma-
tion near the fracture surface (figs. 8 and 9). The frac-
tures of both these specimens appeared to be inter-
granular.
Fractured specimens at 621, 677, and 732°C all
showed apparent transgranular failures at the highest
stresses. As the stress was lowered and the tem-
perature raised, the failures became more inter-
granular, and grain boundary cracking in the speci-
Fig. 9. Grain boundary cracks on specimen tested at 345 men gage length (back from the fracture surface)
MPa at 566°C. Etched with 5 : 1, HCI : HNO3. increased (fig. 10).
182 R.L. Klueh, J.F. King / CZeep and creep rupture of ERhVCr-3 weld metal

Fig. 10. Example of grain boundary cracks and cavities in a specimen tested at 83 MPa at 732°C. Etched with 5 : 1, HCI : HNOs.
(a) Near fracture surface. (b) In gage section back from fracture surface.

4. Diacuaaion tures. We concluded that this difference arose from

the fact that the material tested by Huntington
The only available creep-rupture data for compari- Alloys was in an as-welded condition, whereas our
son are those obtained by Huntington Alloys [4] at material had been tempered 1 h at 732’C after weld-
538, 649, 760, 871, and 982’C. These data consist ing. If the postweld heat treatment created the tensile
of only two or three tests at each temperature, property differences, it did not appear to signifi-
enough data to determine rupture and minimum- cantly affect the creep properties. This assumes,
creep-rate curves. Our data at 566’C fall on the line however, that Huntington Alloys used the same
given by Huntington Alloys at 538’C. At the higher weld metal for both the tensile and the creep tests,
temperatures, our data bracket the Huntington Alloys an assumption that we were unable to verify.
data: our 621 and 677°C data fall on either side of The specimens tested at 5 10 and 566°C were taken
the 649’C data, and the 732’C data fall as expected from welds made with about 7 weld passes on 13-
relative to the 76O’C data. mm-thick plates, while specimens tested at 454,621,
When we compared our tensile data [2] with the 677, and 732OC were from welds made with more
Huntington Alloys data, we found the Huntington than 40 weld passes on 19-mm-thick plates. From the
Alloys material somewhat stronger at all test tempera- tensile studies [2] we concluded that while weld-
 R.L. Klueh, J.F. King / Creep and creep rupture of ERhWr-3 weld metal 183

ment size affects yield strength slightly at elevated Logarithmic creep implies a creep curve with a con-
temperatures, it had little effect on the ultimate ten- stantly decreasing creep rate. Obviously, that was not
sile strength. We further concluded that the only strictly true for the tests in this study, since all the
expected difference in the two weldments should be tests were judged to reach a steady state. However,
due to the dilution of the weld metal by the base the difficulty in determining a steady state for a
metal. Iron from the 2$r-1Mo steel plates was the creep test that does not go into tertiary creep is well
only major dilutent [2]. The dilution of the large known [9]. Several of the low-stress tests at 454’C
weldment was confined close to the interface fall into this category. Similarly, at the low stresses at
between the weld metal and the base metal, while the 510 and 566”C, limited tertiary creep makes the
small weldment contained approximately 10% Fe minimum creep rate uncertain.
throughout. The tensile studies [2] thus indicated no To determine whether eq. (3) applied, we plotted
large dilution effect. Payne [3] indicated that dilu- the strain against log t for the early portion of the
tion by iron should increase the strength, but our low-stress tests at 454, 510, and 566°C (figs. 11 and
results did not indicate strengthening. 12). The two low-stress tests (413 and 434 MPa) at
Although most of the tests at 566°C were made on 454’C indicated logarithmic creep (fig. 11). The
specimens taken from the 13-mm-thick plates, several curve for the 413-MPa test appeared linear to 2500 h,
were also made on specimens from the larger weld- while the curve for the 434-MPa test was linear to
ment. Again, no conclusive difference was noted, about 2000 h, then showed an upward curvature.
though the number of tests involved was small. For Such upward curvature was obvious at 445 MPa at
comparison two specimens from the large weld tested 454’C (fig. 12). Plots of the low-stress tests at 510
at 396 MPa failed much sooner than did one from the and 566°C displayed downward curvature.
smaller weldment. Both were judged to fail prema- While logarithmic creep is confined to a low
turely. Test data from a specimen taken from the homologous temperature range, it is also confined
large weldment and tested at 345 MPa fell on the to low strains. According to Garofalo [8], it occurs to
creep-rupture and minimumcreep-rate curves (figs. 1 strains of about 2 X 10m3. The total “creep” strains
and 2) plotted from test data from speciments taken for the two tests at 454’C were considerably larger.
from the small weldments. On the other hand, a However, early in both these tests a strain burst of
specimen from the small weldment failed prema- about 3% took place. The strains used in fig. 14 were
turely at 345 MPa. Thus, these few tests suggest no creep strains measured after the strain bursts had
dilution effects or other differences that could be
attributed to difference in size of the weldments from
which the specimens were taken.
The creep-curve shapes at the low temperatures are
indicative of logarithmic creep behavior with a
strain-time relation of the type
e=culogttc, (3)
where E is the creep strain, I is the time, and a and c
s t i
are constants that are independent of time. Such a
relation has been found [8] to apply to several metals
and alloys at low homologous temperatures, that is,
at low T/T,,,, where T is the test temperature and T,,,
the melting temperature in K. Garofalo [8] stated
that eq. (3) applied for T/T,,, from 0.05 to 0.3. For 2 1 ’ “““” ’ “““” ’ ’ “““’ ’ ’ “‘U.l
ERNiCr-3, T,,, is about 1700 K and T/T,,, is 0.27, to’ 102 403 104 103
0.30, and 0.33 for 454,510, and 566”C, respectively, TIME (hr)

near the upper limit of the region where Garofalo Fig. 11. Semilogarithmic plot of creep data for ERNiCr-3
indicated eq. (3) applies. weld metal tested at 413 and 434 MPa at 454°C.
184 R. t. Klueh, J.F. Krkg / Oeep and creep rupture of KRMCY-3 weld metal

creep law will be further complicated by the strain

burst phenomenon.
In our tensile studies we concluded that short-
4
range order occurs in ERNiCr-3 weld metal [lo],
This resulted in peaks in the plot of flow stress
against temperature. The peaks occurred in the range
450 to 7OO’C. A local maximum occurred in the
yield strength-temperature relationship near 550°C,
33
while the ultimate tensile strength local maximum
z
occurred near 500°C. An ordered phase, NiaCr, exists
d
f in the binary nickel-chromium system [I 1- 13] with
a critical ordering temperature Tc s 590°C. It was
concluded [ 101 that short-range order due to nickei
2 and chromium gave rise to the observed tensile
behavior. As discussed below, we feel that short-
range order can also explain the creep behavior at low
temperatures and play a role in the strain burst obser-
vations.
Fisher [14] showed that an ahoy with short-
i0' 102 103 to4 $05
range order must be stronger than one without such
TIMEfhrf
order. When a dislocation moves through a partially
Fig. 12. Semilogarithmic plot of creep data for ERNiCr-3
weld metal tested at 455 MPa at 454”C, 345 and 379 MPa at
ordered region, it destroys order. Because more
SlO’C, and 327 MPa at 566’C. energy is required to break the order bonds, it is more
difficult to move a dislocation through a lattice with
short-range order. However, once the first disloca-
occurred. When the strains caused by the strain bursts tion destroys the order, subsequent dislocations can
were ignored, strains of up ot about 5 x 10e3 were move across that disordered slip plane at a reduced
found for the two low-stress tests at 454’C. stress.
Logarithmic creep, therefore, apparently occurs at Logarithmic creep has been explained by exhaus-
low stresses at 454°C. Obviously, more tests at lower tion theories that associated such creep behavior
stresses are required to verify this behavior and deter- with the “exhaustion” of dislocations that can glide
mine the effect of stress on the constants in eq. (3). IS]. Exhaustion is generally attributed to the fact
Because the tests at 5 10 and 566°C as well as those at that a limited number of mobile dislocations are
the higher stresses at 454*C do not follow eq. (3), present when creep begins and are “used up”. How-
some other creep law must apply. The test at 455 ever, exhaustion could also come about by the
MPa at 454°C seems to Iit a parabolic law of the type generation of dislocation barriers within the lattice
that make dislocation motion continuously more
E=pP+&t, (4) difficult.
where /3 and m are constants and P, is the steady-state For the ERNiCr-3 weld metal, we propose that the
creep rate. However, this equation did not fit the low- creep-curve shape observed at 454,5 10, and 566’C is
stress tests at 510 and 5&C. the result of short-range order formation that occurs
Numerous other empirical creep laws have been by diffusion during creep. Short-range order can be
used to fit strain-time data at elevated temperatures visualized as many smaI1 regions conta~ng order
and for large deformations. To determine such a law within the generaliy disordered lattice [ 151. These
for ERNiCr-3 was beyond the scope of the present regions can form and grow during elevated-tempera-
study. That subject will be addressed in later work. In ture exposure (i.e., the degree of short-range order of
addition to the different types of behavior noted at the alloy increases). If this occurs during a creep test,
the different test temperatures, the development of a as more barriers form with time, while others grow, it
 R.L. Klueh, J.F. King / Creep and creep rupture of ERNiCr-3 weld metal 185

becomes more difficult to move dislocations through order dislocation barriers. However, if the stress on
the lattice. Another way to look at this is that fewer the barrier due to a dislocation pileup exceeds some
of the dislocations can move through the lattice critical value, the lead dislocation penetrates the
under the applied stress (exhaustion). Thus, we have barrier and leads to a dislocation avalanche.
logarithmic creep or creep curves of the type Such short-range order barriers can account for the
observed for ERNiCr-3 weld metal at low stresses at decreasing strain burst activity with increasing tem-
454, S 10, and 566°C (fig. 3a). At higher stresses and perature and absence above a certain temperature.
temperatures, the short-range order barriers are not First, the amount of order decreases with an increase
as effective in stopping the dislocations and there are in temperature. The absence of strain burst activity at
deviations from logarithmic creep. Although short- 621’C but occurrence at 566°C may reflect the fact
range order does not cease above T,, the degree of that nickel and chromium are responsible for the
such order should decrease considerably as should the short-range order (for NizCr formation, T, Y S90°C).
effect it has on the properties. As noted, the charac- Although short-range order should not cease above
ter of the creep curves was different at 566 and T,, a somewhat abrupt difference might occur above
62 l°C (Tc 2 S9O”C for Ni&r). and below Tc. Regardless, above some temperature
The strain burst phenomenon was discussed in a the degree of short-range order will decrease until it
separate paper [ 161 and will not be discussed in detail is insufficient to form barriers to continuous dislo-
here. Such obvious explanations as test technique, cation motion.
neck formation, and void or crack formation were With only a few exceptions, all specimens that
eliminated, as were gross microst~~tur~ changes displayed the strain burst phenomenon had bursts
(e.g., recrystallization), since no such changes were during the primary creep stage, while the ones that
observed. The phenomenon is not an effect of the did not failed prematurely. At 454, 510, and S66’C,
cast microstructure of welds, as the effect was also all loading strains exceeded 5% and most exceeded
observed in wrought ERNiC-3 [ 161. 10%. These results indicate that dislocation pileups
Our observations on strain bursts [16] suggested could form during loading and that a strain burst
that deformation is somehow prevented until stress during the primary creep stage subsequently relieved
builds up to a critical level. Then the built-up stress such pileups. The three specimens that did not have
is unlocked, and the specimen suddenly strains. On an strain bursts in the primary stage had significantly
atomic level we could visualize dislocation pileups at greater loading strains than those that did, indicating
some barrier within a given gram. These dislocation that strain bursts probably occurred during loading.
pileups would be in addition to a distribution of Virtually all tests at 4S4’C showed extensive
mobile dislocations that allows for “normal creep”. strain burst activity. Although strain bursts early in
Eventually, the stress buildup on an isolated pileup the primary stage presumably relieved the pileups
forces the lead dislocation through the barrier. Once that occurred during loading, additional pileups had
the lead dislocations breaks through the barrier, to form, leading to strain bursts during the second-
the dislocations behind it are released in a “disloca- ary stage. fn most cases these strain bursts in the
tion avalanche”. The avalanche then triggers the steady-state stage were much smaller than those in
release of pileups in the same grain and in neighboring the primary stage.
grains. The net result is a “deformation band” that The magnitude of the strain bursts during the
moves along the gage section in a manner similar to primary stage for the four lowest stress tests (414,
the movement of a Luders band. A similar result may 434, 455, and 483 MPa) at 4S4’C decreased slightly
occur if certain locked dislocation sources are trig- with decreasing stress. Intuitively, we expect this,
gered, and, in turn trigger nei~bor~g sources in the since the lower the stress is, the less is the extent of
same gram and neighboring grains. pileup formation expected during the loading process.
As was true for the logarithmic creep, the barrier The tests at 455 and 483 MPa also had strain bursts
that can explain such behavior could be formed by during the. steady&ate stage, whereas those at 414
regions of short-range order. That is, exhaustion and 434 MPa did not. However, the latter two tests
creep is the result of the formation of short-range- were discontinued before rupture. Thus, although the
186 R.L. Klueh, J. F. King / Creep and creep rupture ofERNiCr-3 weld metal

pileups formed during loading are released by the stresses lower than those used in this study. As the
strain burst during the primary creep stage, pileups stress is lowered, the magnitude of the strain burst in
again form, The higher the creep stress, the more the primary stage should continue to decrease; that is,
quickly pileups form and are released by a strain as the loading strain decreases, the extent of the
burst. From this we might expect that if the tests at initial pileups should decrease. Eventually, a stress
414 and 434 MPa had been continued, strain bursts should be reached at which strain bursts no longer
would have occurred, since at the lower stress the dis- occur during the primary stage. However, strain
location at the head of the pileup simply takes longer bursts should continue to occur during the steady-
to force itself through the barrier than at higher state stage. This follows because the stress will be
stresses. Only more long-term low-stress tests will too small to move dislocations continuously through
verify this. the barriers.
At 510 and 566’C all but one of the specimens The as-welded ERNiCr-3 was previously shown to
that diplayed strain bursts had the bursts during contain very smail defects scattered through the
primary creep; the exception was a premature failure microstructure, defects of a size that radiography
discussed above. The tests with strain bursts were the could not detect [2]. These defects were always
lowstress tests. These results seem to indicate that at found in grain boundaries and often appeared to be
the highest stresses it is possible to force dislocations grain boundary cracks (fig. 13). The tensile failures
continuously through the barrier. As the stress is occurred transgranularly. However, we found no
lowered, a stress is reached where the dislocations direct evidence to relate the cracks to the grain
cannot be continuously forced through the barrier boundary defects observed on the as-welded material,
and they pile up, and eventually a strain burst although we suspected thet they were connected.
follows. As the temperature was increased and the creep
From the above discussion, we should be able to stress decreased, grain boundary separation became
qualitatively predict the strain burst behavior at more important to the fracture process. On the other

Fig. 13. Example of scattered cracks or voids that were observed in as-welded ERNiCr-3. Etched with 5 : 1, HCI : HN03.
Fig. 14. Photomicrographs that demonstrate the cause of “knobby” deformation during creep. (a) Adjacent to position where several grains meet. (b) Adjacent to
grain boundary cracks. (c) Enlarged view af crack in 0). (a) and (c) were etched with 5 : 1, HCI : HNOa.
188 R.L. Klueii, 3.F. King / cleep and creep rupture ofERM@3 weld metal

hand, at all stresses at 454’C and at all but the lower summarizes the results and conclusions:
stresses at 510 and 566’C, the failures were trans- (1) Metallography revealed scattered grain
granular. Nevertheless, as was true for the tensile boundary cracks or voids, which radiography could
tests, there were again scattered instances of grain not detect.
bounda~ separation. And again, we must logically (2) Creep behavior fell into two temperature
suspect a connection between these grain boundary regimes, which respectively consisted of tests at
cracks and those seen in the as-welded microstruc- 454, 510, and 566’C and tests at 621, 677, and
ture. 732’C. Although the 621’C test results generally
The “knobby” deformation in tensile tests was resembled those at 677 and 732’C, their characteris-
previously [2 ] related to the microstructure of the tics often fell between those of the low- and high-
weld. Similarly, in creep specimens met~o~ap~c~y temperature regimes.
examined (fig. 14), knobs often appeared on the sur- (3) The creep-rupture curves in the low-tempera-
face near the position where several grain boundaries ture regime were quite flat -especially at high
met (fig. 14a). In other cases the knob resulted from stresses -and were characterized by a change in
grain boundary cracks in material adjacent to the slope. Similar behavior was noted for logarithmic
external surface. plots of stress against minimum creep rate. Stress
The connection between strain bursts and prema- exponents from the latter plots were in the range of
ture failures may well lie in the origin of the knobby 25 to 30 for the low-stress tests for the low-tempera-
deformations. If during a large strain burst, an ture regime, as opposed to 8 to 10 for the high-
instability - a knob, which is really the start of a temperature tests.
neck - develops, it could lead to failure.. This is (4) Creepcurve shapes differed for the two tem-
especially true if that instability results from a grain perature regimes. At the low temperatures, creep
boundary crack such as that shown in fig. 14b. Neck curves were characterized by a high initial creep rate
formation is favored by high strain rates and low tem- followed by a rapid transient to the steady state, and
peratures, and the strain rates during a strain burst essentially no tertiary creep. Creep curves for tests in
would appear to be quite large. Hence, if a cavity is the high-temperature regime displayed a lower initial
present in conjunction with the neck (knob), defor- creep rate, less steady-state creep, and a longer period
mation can then proceed to failure, turning the creep of tertiary creep.
test into a tensile test. At 621,677, and 732”C, there (5) For certain tests at 454, 510, and 566”C, the
were knobby deformations only for a few, of the creep curves exhibited strain “jumps” termed “instan-
higher stresses and no premature failures. It should be taneous elongations” or “strain bursts”. Such strain
noted that although the premature failures may be bursts were observed during primary creep and/or
due to the interaction of strain bursts with features during the steady-state stage. The percentage of tests
that are specific to the weld metal cast structure that displayed such strain bursts increased with
(i.e., knobby deformation and scattered grain decreasing temperature: that is, 40% of tests at
boundary cracks), the strain bursts are not caused by 566’C, 67% at 5 lO”C, and 73°C at 454’C exhibited
the cast structure because they also occur in wrought them. No strain bursts were observed above 566°C.
ERNICr-3 where there is no knobby deformation or (6) Several tests at 454, 510, and 566’C ruptured
premature failure [ 161. prematurely; that is, the tests did not obey the
rupture curves established by the majority of tests.
Evidence suggested that the premature failures were
5. Summery and conclusions related to both the strain burst phenomenon and the
scattered grain boundary cracks that were observed
Creep-rupture tests were made pn ERNiCr-3 by optical microscopy in the as-welded alloy.
(Inconel 82) weld metal over the range 454 to 732*C. (7) We postulated that the strain burst phenome-
Tests were made on weld metal specimens taken from non is connected with the fo~ation of dislocation
weldments made with the automatic gas tungsten-arc pileups within the matrix. The localized release of a
process with cold-wire filler additions. The following pileup triggers neightboring pileups and gives rise to
 R.L. Klueh, J.F. King / Czeep and creep rupture of ERNiO-3 weld metal 189

the propagation of a Luders-type deformation band References

along the gage length. The pileups were concluded to
be caused by the formation of short-range order in [l] J.F. King, M.D. Sullivan and GM. Slaughter, Weld. J.
the alloy. Short-range order also gives rise to the dif- (Miami) 56 (1977) 354-9.
[2] R.L. Khreh and J.F. King, Weld. J. (Miami) 59 (1980)
ferent creep characteristics in the low-temperature
114-9.
regime. [3] B.E. Payne, Met. Con&r. Br. Weld. J. 1 (1969) 79.
(41 S. Keiser, private communication.
[S] F. Garofalo, R.W. Whitmore. W.F. Domis and F. von
Acknowledgements Gemmingen,Trans. Met. Sot. AIME 221 (1961) 310.
[6] W.E. Leyda and J.P. Rowe, American Society for Metals
Tech. Report (USA) P96.1 (1969).
We wish to acknowledge several people who were [7] R.L. Klueh, J. Nucl. Mater. 79 (1979) 363.
instrumental in completing this work: J.L. Griffith [8] F. Garofalo, Fundamentals of Creep and Creep-Rupture
conducted the creep tests; C.W. Houck carried out in Metals (Macmillan, New York, 1965) p. 159.
the metallography; D.P. Edmonds, V.K. Sikka, C.R. [9] DA. Woodford, Mater. Sci Eng. 4 (1969) 146.
[lo] R.L. Kluehand J.F. King, Met. Trans. 10A (1979) 1543.
Brinkman, and GM. Slaughter reviewed the manus- [ 111 N.V. Semenova, Phys. Met. Metallogr. (Engl. Transl.)
cript; S. Peterson edited the manuscript; and A.F. 6(b) (1958) 57.
Rice prepared the final version of the manuscript. [12] H. SchulIer and P. Schwaab, Z. MetaJlk. 1 (1960) 81.
[ 131 RX Davies, J. Phys. Chem. Solids 24 (1963) 985.
[14] J.C. Fisher, Acta Met. 2 (1954) 9.
[15] J.B. Cohen and ME. Fine, J. Phys. Radium 23 (1962)
749.
[16] R.L. KJueh and J.F. King, Scripta Met. 13 (1979) 205.