Sunteți pe pagina 1din 91

APPENDIX A

Final Report
The University of Tennessee
Knoxville

Professor John D. Landes


Dr. Kang Lee

November 1998

Final Report on Computer Controlled Microindenter System


This is the final report for the University of Tennessee subcontract on "Study of the
Application of Microprobe Methodology to Material Response". It covers the period December
1,1996 to December 31, 1997. The work on this subcontract was conducted by John D. Landes,
Professor of Mechanical and Aerospace Engineering and Engineering Science (MAES) and
Kang Lee, Postdoctoral Associate in MAES.

During this period work was conducted on four topics:


1. Literature survey
2. Fatigue testing
3. Simulation analysis of indenter response
4. Numerical simulation of fatigue testing

The following contains a summary of the results in each area.

1. Literature survey

A computerized literature search was conducted at the main library of the University of
Tennessee, Knoxville (UTK). The object was to gather information on the state-of the-art models
and correlations between fatigue damage and associated changes in the stress-strain behavior of
materials, to study the results of indentation measurements on various material systems and to
study the effects of residual stresses on the uniaxial and multiaxial deformation of aluminum and

steel alloys. Two data bases were used in the searches, a compendex from UTK web-based
databases using ERL/Web SPIRS and an Internet Database Service (IDS) from Cambridge
Scientific Abstracts Web Service. The main part of this study was conducted in the first half year
of the program and the results were reported in the previous semi-annual report. The conclusion
from the literature survey was that no one had previously established a good quantitative
correlation between the stress-strain response and the amount of fatigue damage although it is
clear that a qualitative correlation exists between the two.

2. Fatigue testing and results

The fatigue testing which was in its initial stages last reporting period has been completed
during the past six months of the program. Two materials were tested, a 2024-T4 aluminum alloy
and a 7075-T6 aluminum alloy. The first alloy, 2024-T4, is one that exhibits extensive cyclic
hardening during the fatigue test; that is, the stress-strain curve increases during cyclic loading
relative to the monotonic stress-strain curve. The second alloy, 7075-T6, shows almost no
change in stress-strain properties during the cyclic test. A flat tensile geometry was used for the
fatigue testing rather the typical round low cycle fatigue geometry so that the specimens can be
directly indented without having to deal with the curvature problem that would be introduced by
a round specimen. This specimen is recommended as an alternative to the round fatigue
specimen in the ASTM Method E 606, "Standard Recommended Practice for Constant
Amplitude Low Cycle Fatigue Testing". The fatigue specimen geometry used is shown in Fig. 1.
The tests were conducted on a 22 kip closed loop servo-hydraulic machine. The test was
conducted in displacement control; this was close to strain control but was not as dangerous.

Initially the strain versus life behavior of the material was measured so that subsequent
evaluations of indenter behavior could be made as a function of elapsed life. After the strain
versus life relationship was established, interrupted tests were conducted so that ABI tests could
be made on specimens that had been cycled to a fraction of the failure life and also so that
monotonic tensile tests could be conducted on specimens that were cycled to a fraction of the
failure life. In addition fatigue tests were conducted on two specimens that had been subjected to
multiple indentation tests prior to fatigue testing. This was to demonstrate the nondestructive
nature of the indention test.
A summary of all of the fatigue and tensile tests conducted is given in Table 1. The
2024-T4 aluminum alloy has the code AL2 and the 7075-T6 aluminum alloy has the code AL7.
The results of the 2024-T4 and 7075-T6 aluminum fatigue tests are given in Figs. 2 through 10.
For low cycle fatigue the main result is the cyclic life, measured as strain range, a versus life,
cycles to failure, Nf. The plotted strain range, a, is half of the total strain range. The test is
conducted in completely reversed strain control so a represents the total positive strain range or
the total negative strain range. The cycles to failure are plotted as 2Nf so that 2Nf = 1 gives a
relationship between a half cycle during the fatigue test and a monotonically loaded test. The
strain range, a, is made up of an elastic component, ea and a plastic component, pa.
a = ea + pa

Each component of strain is plotted as a function of 2Nf on log-log plots. The elastic and plastic
components are fitted with straight lines on the log-log plots. These have the form

ea =

' a
( 2 N f ) b + ' f ( 2n f ) c
E

where 'a, 'f, b and c are fitting constants and E is the elastic modulus. When these fits are
combined as given in eq. (1), the total strain range, a, versus life, 2Nf, has the form
3

a =

' a
( 2 N f ) b + ' f ( 2 N f ) c
E

The strain versus life plots for the 2024-T4 aluminum are given in Fig. 2a and for the 7075-T6
aluminum in Fig. 2b. The two are compared on Fig. 2c.
A second way that the low cycle fatigue data are treated is to plot the cyclic stress-strain.
For this the half stress range, a, is plotted as a function of the half strain range, a. For materials
that change the stress-strain property during cycling, hardening or softening, the cyclic
stress-strain is plotted near the half life, 0.5 Nf with the hope that the hardening or softening
reaches an equilibrium called saturation, a point where there is no more increase in the
stress-strain curve. The 2024-T4 had extensive hardening and did not completely saturate at half
life. Therefore the cyclic stress-strain was plotted at 50% life, 0.5 Nf and 90% life, 0.9 Nf. The
2024-T4 cyclic stress-strain curve is plotted in Fig. 3a. From this it is obvious that the material
2024-T4 continues to harden during entire fatigue life but experiences most of the hardening
during the first half of life. The 7075-T6 cyclic stress-strain curve is plotted in Fig. 3b. It does
not show much hardening.
One of the objectives of the fatigue testing was to see if the indenting of the specimen
would have any effect on the fatigue life. Since it is stated that an indenter test is non-destructive,
it is important to verify this for fatigue behavior. Fatigue tests were conducted on two 2024-T4
aluminum alloy specimens with five indentations each. The median strain life curve for
specimens tested in the original condition is compared with the tests that were subjected to
Automatic Ball Indention (ABI) before fatigue testing in Fig. 4. It can be seen that these
specimens appear to have slightly longer life than the specimens tested in the original condition.
Considering typical scatter it is probably better to say only that the result of the tests with ABI

pretest did not show any adverse effects and the ABI is completely non-destructive in terms of
fatigue damage.
Another objective of the fatigue testing was to look at the cyclic stress-strain changes
during the test to see if this change could be an indicator of the fatigue damage. To do this first
the typical hardening behavior of the 2024-T4 aluminum is given in a series of figures that show
the load and displacement values as measured during the test at various percentages of Nf. The
load versus displacement curves for a 2024-T4 specimen subjected to 0.0073 strain are given in
Fig. 5. A composite of the cycles is in Fig. 5a. The specimen appears to harden, that is, the load
increased at constant displacement range as the test progressed. To see this more clearly the
result from cycle 10 is plotted separately with the result from cycle 100, Fig. 5b, with cycle 390,
Fig. 5c, and with cycle 700, Fig. 5d. The total cyclic life for this specimen was 772 cycles. The
same type of plot is made for a 7075-T6 specimen in Fig. 6a through 6d. This alloy did not show
much change in the load because it is not a hardening material. It has a neutral response to cyclic
loading, neither hardening nor softening.
Several specimens were cycled to a certain percentage of the life and then a monotonic
stress-strain curve was conducted. The monotonic stress-strain behavior for the various levels of
cyclic loading is shown in for 2024-T4 in Fig. 7 and for 7075-T6 in Fig. 8. In Fig. 7 it is obvious
that the stress-strain response for the 2024-T4 is increased with increasing cycles. After 100
cycles, which is a little more than 10% of life the majority of the hardening already has occurred,
but at 400 cycles, which is more than 50% of life there is some additional hardening. This
material continues to harden somewhat throughout its entire life. This hardening could provide a
measurement of the amount of fatigue damage that has occurred. From Fig. 8 is apparent that
the 7075-T6 does not change its stress-strain response very much as a result of cyclic damage

and therefore would not provide a way to use the stress-strain changes to measure the amount of
fatigue damage. A further objective was to see whether the ABI test could measure the
stress-strain changes that accompany cyclic loading. The ABI test measures the true stress-true
strain result rather than the engineering stress and strain. To make comparisons between the ABI
test result and the monotonic test result all tests were converted to true stress-true strain. The
result of the ABI stress-strain measurements for the 2024-T4 is shown in Fig. 9. This includes
the stress-strain curves measured at 0, 70 and 400 cycles. It is clear from this result that the
hardening resulting from the cyclic loading can be measured but that this nearly saturates with
cyclic loading. The comparison between the standard tensile test and the ABI test is made in Fig.
10a for all of the test results on 2024-T4. Fig. 10b shows a comparison of the tensile results, that
is the tensile test versus ABI test stress-strain measurement, on virgin material. Fig. 10c shows
the comparison after 71 cycles and Fig. 10d shows the comparison after 400 cycles. The standard
tensile test and the ABI test follow the same hardening trends.

3. Numerical Simulation of Indenter Response

A finite element simulation of the indenter response on a material was conducted to see if
it is possible to predict the indenter behavior during the loading cycle. With a successful
prediction of a test result, the finite element method could be used to study the effect of various
indenter variables on material response. This could include the indenter size, profile, material,
application of multiple indenters and spacing. In addition the local details of the stress and
deformation response of a material could be predicted in a way not possible from test results.
The first step in the analysis of the indenter response was to see if a typical loading response

record from a test conducted with the Microprobe system at Advanced Technology Corporation
Headquarters could be successfully modeled by the finite element system.
The original system chosen for the simulation study was an A533B steel plate which was
indented with a 0.062 inch ball geometry of tungsten carbide. The stress-strain analysis is treated
as a two-dimensional axisymmetric problem in which half of a diametrical cut through the center
of the indenter ball and the plate is analyzed. The pre and post processing code used to prepare
the finite element mesh was PATRAN 5.1. The finite element analysis code was ABAQUS
5.5-1. These software packages were run on a SUN workstation. The element type used was a
Quad 4, 2-D solid axisymmetric element. The indenting simulation is modeled as a non-linear
contact problem. An overall analysis with results were given in the previous progress report. This
included a convergence analysis and the stress and displacement predicted by the simulation
during the indentation process.
The first simulation analysis of the indenter response used nominal dimensions of the
indenter, rod, ball and brazed attachment system. To conduct a more precise analysis the exact
dimensions of the indenter were measured with a calibrated traveling microscope so that the
simulated indentation procedure could be predicted based on actual dimensions. This microscope
can measure the dimensions to 0.0005 in. accuracy. A summary of the measurements for nine
indenters covering five different sizes is given in Fig. 11 along with a schematic of the indenter
head showing the definitions of the various symbols used to identify the measurements.
Two of the indenter sizes were chosen to conduct a simulated indentation modeling. They
were the 0.030 inch and 0.062 inch heads diameters. The overall finite element model of the
indenter 0.030 indenter head is shown in Fig. 12a. The detailed model near the indenting surface
is shown in Fig. 12b. The material being indented in this second simulation is the 2024-T4

aluminum alloy used in the fatigue tests. For this material accurate stress-strain data is available
from tensile tests. It was used for the indenter testing so that direct comparisons can be made
between the indenter simulation and the actual indenter test.
The deformation and stress distribution for the 0.030 inch indenter head at 78.5 lb load is
given in Fig. 13a. The same distributions after unload of the 78.5 lbs is given in Fig. 13b. The
stress distributions shown are the von Mises equivalent stresses. These values more accurately
indicate the stress value relative to a yielding condition. After the 78.5 lbs has been unloaded,
Fig. 13b, the stress pattern represents the residual stresses that are remaining as a result of the
indentation.
The purpose of the finite element simulation in this section was to see if the ABI test
could be simulated numerically. A comparison between the actual ABI test and the simulation of
the test is shown in Fig. 14. Fig. 14a has the load versus depth, measurements from ABI and
predictions. Fig. 14b shows the plastic component of both test and simulation. For this indenter
head size the total displacement has a slight but constant difference but the plastic component of
the displacement matches exactly. The difference in total displacement match is in the order of
2x10-4 inch.
The finite element model for the simulation using a 0.062 inch indenter head is shown in
Fig. 15. Fig. 15a has the entire indenter and Fig. 15b has the details near the indenter head. The
deformation and von Mises stress distribution is shown for a 319 lb load in Fig. 16a and for the
residual pattern after unload from 319 lb in Fig. 16b. The larger diameter indenter requires a load
increase of about the difference in area, that is, the ratio of indenter diameters squared. The load
versus depth of penetration for the ABI test and the simulation is shown in Fig. 17; Fig. 17a for
total depth and Fig. 17b for plastic depth. For this case the plastic displacements also have a

small but constant difference.


For both indenter sizes the results of the simulation match the test results very well. This
would suggest that a simulation of a test condition can predict actual behavior within a fairly
good accuracy. In the future when the parameters of the indenter system might be changed, a
simulation could be first conducted to see if the desired result is obtained. Things like changing
indenter head geometry or material could require an expensive trial and error sequence to
determine optimum test conditions. Each new head could require an expensive material purchase
and machining operation. With the simulation the parameters can be easily changed without the
need for costly machining or material purchases.

4. Numerical Simulation of Fatigue

To calibrate the instrumentation for the fatigue tests and to see that the results are
acceptable, a simulation analysis of the fatigue test was performed numerically. The simulation
was conducted with a finite element model that used the same software as the simulated indenter
analysis. The pre and post processing code used to prepare the finite element mesh was
PATRAN 5.1. The finite element analysis code was ABAQUS 5.5-1. These software packages
were run on a SUN workstation. The simulation measured the stress response to an applied strain
level; this was applied as a displacement placed over the gripping area of the fatigue specimen in
a manner that simulates the actual loading during the fatigue test. The total maximum strain level
was positive 0.02. The specimen was then unloaded to the same strain level in the negative
direction, a strain of negative 0.02. The cycling was continued by again increasing to the same
positive level of strain and unloading to the same negative strain level. The entire simulation

took approximately 20 hours of computer time to run. During each cycle the stress and
deformation were measured so that a picture of the deformation and the stress changes could be
recorded. The effect of the cyclic loading on the specimen model is shown on a series of figures
showing the deformed grid after a set number of cycles.
The material simulated in fatigue was the 2024-T4 aluminum alloy. The monotonic
stress-strain properties measured in a tensile test were used to relate the strain to the stress. A
model on the finite element mesh is shown in Fig. 18. The deformation response is shown in Fig.
19a through 19e. These plots correspond to cycles 1, 20, 40, 60 and 80. Each plot shows two
models, one for the deformation at the positive 0.02 strain level and one at the negative 0.02
strain level. It is possible to see in these plots a progressive shape change caused by the cyclic
loading. The original specimen shape is shown as a faint background outline. Although the actual
test does not exhibit the dimensional change shown in the simulation, the simulation does
indicate a progressive damage in the fatigue test. The actual test specimen failed before 80 cycles
at this strain level.
The stress response is shown in Figs. 20a through 20f. The stresses shown are first the y
direction stress, yy, and second the von Mises equivalent stress. For example, Fig. 20a shows yy
for cycle 1 and Fig. 20b shows the von Mises stress for cycle 1. The same pattern is repeated for
cycle for cycle 60, Figs 20c and d, and the for cycle 80, Figs. 20e and f. The stress pattern,
particularly, yy, is uniform at cycle 1 and reaches a maximum of about 65 to 70 ksi. After 60
cycles the stress pattern segregates into regions of high and low stresses with the highest regions
reaching the range of 105 to 110 ksi. At cycle 80 the small regions of high stress reach even
higher maximum values. These regions of high stress occur in the parts of the specimen that

10

forms sharp corners as a consequence of the fatigue deformation. The overall increases in the
local stresses correspond to the cyclic hardening character of the material.
To show how the fatigue simulation predicts the cyclic hardening behavior of the
material, the cyclic load versus displacement from the simulation test is shown in Fig. 21 and the
cyclic stress-strain for the simulation is in Fig. 22. In both cases a cyclic hardening behavior was
predicted. This prediction comes from the simulation model itself, not from a change of the
stress-strain property used in the model. The original monotonic stress-strain behavior of the
material was used for the entire simulation.
The fatigue simulation does not predict and actual failure point but does show that the
continued cycling causes severe deformation changes. These deformation patterns simulate the
formation of stress concentration areas that could lead to the formation of a crack and subsequent
failure of the specimen. In addition the stresses which are fairly uniformly distributed at the first
cycle, become segregated into high and low stress regions with continued cycling. The maximum
stresses in these regions increase with increased cycling. The effect of fatigue test simulation is
interesting and should be studied further to understand the full implications of the result.

11

12

13

Fig. 1 Low-Cycle Fatigue Test Specimen Configuration

14

Fig. 2a Strain Amplitude versus Fatigue life for 2024-T4 Aluminum

15

Fig. 2b Strain Amplitude versus Fatigue Life for 7075-T6 Aluminum

16

Fig. 2c A Comparison of 2024-T4 and 7075-T6 Aluminum Srain-Life Plots

17

Fig. 3a Cyclic Stress-Strain Relationship From Low Cycle


Fatigue Test for 2024-T4 Aluminum

18

Fig. 3b Cyclic Stress-Strain Relationship From Low Cycle


Fatigue Test for 7075-T6 Aluminum

19

Fig. 4 Strain Amplitude versus Fatigue life for 2024-T4 Aluminum,


Non-ABI Data vs. Pre-ABI Points Data

20

Fig. 5a Typical Cyclic Load versus Displacement Curve for 2024-T4


Aluminum, Specimen AL209, All Cycles

21

Fig. 5b Typical Cyclic Load versus Displacement Curve for 2024-T4


Aluminum, Specimen AL209, Cycle 10 vs. Cycle 100

22

Fig. 5c Typical Cyclic Load versus Displacement Curve for 2024-T4


Aluminum, Specimen AL209, Cycle 10 vs. Cycle 390

23

Fig. 5d Typical Cyclic Load versus Displacement Curve for 2024-T4


Aluminum, Specimen AL209, Cycle 10 vs. Cycle 700

24

Fig. 6a Typical Cyclic Load versus Displacement Curve for


7075-T6 Aluminum, Specimen AL705, All Cycles

25

Fig. 6b Typical Cyclic Load versus Displacement Curve for


7075-T6 Aluminum, Specimen AL705, Cycle 2 vs. Cycle 10

26

Fig. 6c Typical Cyclic Load versus Displacement Curve for


7075-T6 Aluminum, Specimen AL705, Cycle 2 vs. Cycle 155

27

Fig. 7 Monotonic Stress-Strain Curve from Standard Tensile Test


for 2024-T4 Aluminum at Different Precycle Stages

28

Fig. 8 Monotonic Stress-Strain Curve from Standard Tensile Test


for 7075-T6 Aluminum at Different Precycle Stages

29

Fig. 9 Monotonic True Stress-True Plastic Strain Curves from


ABI Tests for 2024-T4 Aluminum at Different Precycle Stages

30

Fig. 10a A Comparison of True Stress-True Strain Curves from Standard


Tensile Test and from ABI Test, 2024-T4 Aluminum, All Curves

31

Fig. 10b A Comparison of True Stress-True Strain Curves from Standard


Tensile Test and from ABI Test, 2024-T4 Aluminum, All Virgin Materials

32

Fig. 10c A Comparison of True Stress-True Strain Curves from Standard


Tensile Test and from ABI Test, 2024-T4 Aluminum, At About 15% Fatigue Life

33

Fig. 10d A Comparison of True Stress-True Strain Curves from Standard


Tensile Test and from ABI Test, 2024-T4 Aluminum, At About 65% Fatigue Life

34

35

Fig. 12a A Finite Element Simulation Model of


ABI Test with 0.030 in. Indenter Head

36

Fig. 12b A Finite Element Simulation Model of ABI Test with 0.030 in.
Indenter Head, Details Near the Indenter Head Tip

37

38

39

40

Fig. 14a Load Versus Depth from Finite Element ABI Test Simulation,
Comparing with ABI Test Result, 0.030 in. Indenter Head

41

Fig. 14b Load Versus Plastic Depth from Finite Element ABI Test
Simulation, Comparing with ABI Test Result, 0.030 in. Indenter Head

42

Fig. 15a A Finite Element Simulation Model of


ABI Test with 0.062 in. Indenter Head

43

Fig. 15b A Finite Element Simulation Model of ABI Test with


0.030 in. Indenter Head, Details Near the Indenter Head Tip

44

45

46

Fig. 17a Load Versus Depth from Finite Element ABI Test Simulation,
Comparing with ABI Test Result, 0.062 in. Indenter Head

47

Fig. 17b Load Versus Plastic Depth from Finite Element ABI Test
Simulation, Comparing with ABI Test Result, 0.062 in. Indenter Head

48

Fig. 18 A Finite Element Simulation Model of Fatigue Test

49

Fig. 19a A Deformation Plot from Finite Element Fatigue


Test Simulation, 2024-T4 Aluminum, Strain = 0.02, Cycle 1

50

Fig. 19b A Deformation Plot from Finite Element Fatigue


Test Simulation, 2024-T4 Aluminum, Strain = 0.02, Cycle 20

51

Fig. 19c A Deformation Plot from Finite Element Fatigue


Test Simulation, 2024-T4 Aluminum, Strain = 0.02, Cycle 40

52

Fig. 19d A Deformation Plot from Finite Element Fatigue


Test Simulation, 2024-T4 Aluminum, Strain = 0.02, Cycle 60

53

Fig. 19e A Deformation Plot from Finite Element Fatigue


Test Simulation, 2024-T4 Aluminum, Strain = 0.02, Cycle 80

54

55

56

57

58

59

60

Fig. 21 The Cyclic Load Versus Gage Line Displacement from Finite Element
Fatigue Test Simulation, 2024-T4 Aluminum, Strain = 0.02

61

Fig. 22 The Cyclic Stress-Strain Curves from Finite Element Fatigue


Test Simulation, 2024-T4 Aluminum, Strain = 0.02

APPENDIX B

Final Report

Professor K. Linga Murty


North Carolina State University
November 1998

North Carolina State University


Department of Nuclear Engineering
College of Engineering
Box 7909 Raleigh NC 27695-7909
Prof. K. Linga (KL) Murty
(919) 515-3657 FAX 515-5115

Study of Solder-Alloys
As a part of the DoD-SBiR-II (Contract No. N00421-96-C-1121 to the Advanced
Technology Corporation), NCSU group was to investigate and demonstrate the capabilities of
SSM based on ABI technology in characterizing the --T relationship for various lowtemperature alloys commonly used for solder applications in microelectronics. To this end, we
have considered Sn5%Sb, pure lead, Pb5%Sn and Pb62%Sn which comprise various classes of
materials from the deformation/creep constitutive equation point-of-view: class -M, class-A and
superplastic. First type of testing included ABI tests under constant indenter velocity with
velocity varied by several orders of magnitude. Second type of testing included load relaxation
tests using cylindrical indenter so that constant load corresponds to constant stress. Third type of
tests that were contemplated in the beginning were the creep tests under constant stress/load
again first using cylindrical indenter and then followed by ball indenters. However, due to the
change in the time and cost (reduced by about 50%), we have not considered the third type here
which involves development of new algorithm and time consuming tests. Notwithstanding this
cut-back, we demonstrate from the first two types of test methods, the viability of SSM in
obtaining the desired information. In addition to tests at ambient, we have performed at elevated
temperatures to obtain activation energy for creep/deformation in Sn5%Sb while the variations
of with were also characterized for the Pb-alloys at an additional temperature. The results
need further analyses which involves more funding and time; however, as mentioned earlier,
these results do demonstrate the capability (this being the major objective of the substudy).
Sn5 %Sb
Tin-antimony alloy with chemical composition Sn-5%Sb was obtained in the form of 1 mm
thick sheet from Alpha Metals, Inc. Tensile creep specimens with a gauge length of 12.4 mm
were prepared from the sheet parallel to the rolling direction. Constant load creep tests were
performed at 298, 373, 423 and 473 K at stresses varying from 1.25 to 30 MPa, and the creep
strains were measured using LVDT.

KL Murty

page 1

Automated ball indentation tests were carried out at 298, 323, 373 and 423 K. For ABI tests, no
specific specimen preparation was required except that the specimen should be flat and the
surfaces must be parallel. A light mechanical polish (600 grade polish) was given in order to
ensure that surface contamination did not influence the ABI test results and also to achieve flat
parallel surfaces. ABI tests involve pushing a spherical indenter into a specimen while
monitoring the load versus depth of penetration. Tests were carried out using a table top
Stress-Strain Microprobe (SSM) System, Model PortaFlow-Pl (patented by the Advanced
Technology Corporation, Oak Ridge, TN). A tungsten carbide ball indenter of 1.575 mm
diameter was used and the maximum depth of penetration of the indenter was about 0.1 mm; the
corresponding diameter of the indentation was about 0.375 mm. Each indentation covered a large
number of grains since the grain size of the material was small ( 10 m).
The creep curves were generally characterized by normal primary creep, short steady-state
regime and relatively long tertiary region as shown in Fig. 1. This was typical of all the test
conditions. The minimum creep rates were evaluated from the creep curves, and Figure 2 shows
the variation of minimum creep rate with stress at different temperatures. A power law
relationship of the form,
= A n , (1)
was found to be obeyed over the entire stress range at 298, 373 and 423 K with a value of n5.
At 473 K, a transition occurred from a value of n=5 in the high stress region to a value of n=3 in
the low stress region (below 2 MPa). These results are similar to the tensile test data which also
indicated the transition at 473 K.1

Figure 1
Typical creep curves (sn5%Sb)
1

R. K. Mahidhara, K. L. Murty and F. M. Haggag, Creep and Mechanical Properties of Sn5%sb Solder, in Design

and Reliability of Solders and solder Interconnections, R. K. Mahidhara et. al., ( 1997) 75.

KL Murty

page 2

Figure 2
Minimum creep rate vs stress at various temperatures
The activation energy for creep was determined in the high stress (n=5) region, from an
Arrhenius plot of ln(strain-rate) against reciprocal of absolute temperature (Figure 3). Least
squares analysis of the data yielded a value of Q=12.61.1 kCal/mole. As reported earlier, the
tensile test data yielded much lower values (~6.4 kCal/mole).

Figure 3
Activation Energy Plot (Creep Data)
ABI Results
From previous studies, it has been well established that ABI can be employed for studying the
stress-strain behavior of materials and that the results agree very well with conventional tensile

KL Murty

page 3

results. ABI is a relatively simple, rapid and non-destructive technique and requires small
amounts of material with minimum material preparation. The principle of the test technique and
the calculation procedures have been reported previously. In the present investigation, ABI tests
were carried out at 294, 323, 373 and 423 K over a range of indenter velocities (strain-rates). The
primary data generated in an ABI test is the indentation load versus depth of penetration curve,
and all the relevant mechanical properties are evaluated by further analysis of this data.
Indentation load versus depth of penetration curves for Sn5%Sb obtained at two different
indenter speeds are shown in Fig. 4. Single cycle method is used in lieu of the standard partial
unloading technique because of the relatively fast rates employed in characterizing the strain-rate
variation of flow behavior. The corresponding true stress and true strain values are calculated by
the system software.

Figure 4. Typical indentation load versus depth curves at various indenter speeds.
The true plastic strain is determined from the equation,
d
, (2)
p = 0 .2 p
D

where dp and D are the plastic indentation diameter and diameter of indenter respectively. The
corresponding flow stress is calculated from the following equation,

4P
d p2 , (3)

where x is the true indentation stress and x is a parameter which depends among other things also
on the system compliance. With ABI technique, yield stress, tensile strength, and the work
hardening parameters of the material are evaluated directly from the flow curve. The flow curve
is usually represented by the equation,
= Kn , (4)
where K and n are the strength coefficient and strain hardening exponent respectively. From the
fact that the strain hardening parameter n is equal to the true uniform strain (u) at the ultimate
tensile stress of the material, the true ultimate tensile stress can be evaluated as,

KL Murty

page 4

TS = Knn . (5)
The ultimate tensile strength (UTS, the engineering value) is then obtained from the
equation,
n n
. UTS = K { e } (6)
The indenter velocity (vi) is converted into tensile strain rate using the relationship,9

2 vi
5 dp

(7)

Figure 5 shows the variation of strain rate with true ultimate tensile stress at these temperatures.
A power law relationship was found to be obeyed with a value of n=5.4 which agrees closely
with that determined from creep experiments at high stresses. From the Arrhenius plot (Fig. 6),
the activation energy for deformation was determined to be 13.11.8 kCal/mole corresponding to
a stress level of 60 MPa which is also in good agreement with that for creep. The values of n and
Q obtained from both ABI and creep point out to the same deformation mechanism which we
propose to be viscous glide as described below.

Figure 5
Variation of strain rate with UTS from ABI tests

Figure 6
Arrhenius plot ABI Data

SSM is a versatile tool in the sense that apart from stress-strain properties, it is capable of
evaluating stress relaxation behavior of the material. Stress relaxation tests offer the possibility
of studying the deformation behavior over a wider range of strain rates compared to conventional
creep tests. It is clear from the study that by decreasing the ball diameters to micron-size, SSM
can be adopted for in-service or in-situ application for characterizing the solder-condition
leading to the predictability of their remaining life.

KL Murty

page 5

Deformation Mechanisms (SnS%Sb)


Creep deformation mechanisms are identified mainly by the value of the stress exponent n and
the activation energy Q.2, 4 Dislocation creep occurs as a result of glide and climb of edge
dislocations.2 These are sequential processes so that the slower of the two acts as the rate
controlling step. Dislocation creep in pure metals is characterized by a value of n=5, and climb of
edge dislocations being slower, becomes the rate controlling mechanism. Solid solution alloys
are broadly categorized as class-M alloys if the value of n is equal to 5, and class-A alloys if the
value of n is equal to 3.3 In class-M alloys, the rate controlling mechanism is the climb of edge
dislocations just as in the case of pure metals. In class-A alloys, the rate controlling mechanism
is solute-diffusion controlled dislocation glide since the dislocations are locked by solute
atmospheres such that the rate of glide becomes slower than that of climb. Such a mechanism
which is also known as microcreep4 predicts an activation energy for creep which is equal to that
for diffusion of solute elements. In either case, recovery takes place by the climb of dislocations.
On the basis of the n value in the low stress region at 473 K, creep deformation is perceived to be
controlled by viscous glide of dislocations.9 The activation energy for deformation would be
equal to that for solute diffusion via vacancy mechanism. Since the low stress behavior has been
exhibited only at one temperature, the activation energy for the mechanism could not be
evaluated.
Assuming that the low stress region is in fact controlled by viscous glide of dislocations, one
would expect the dislocations to break away from the solute atmospheres as high stresses are
applied.5 Hence, at higher stresses, climb would be the rate controlling mechanism with a value
of n5. It may hence be concluded that in the present study the rate controlling mechanism is
dislocation climb at all the test conditions except at very low stresses at 473 K (Fig. 3). As
Murty17 has shown in the case of an Al-Mg alloy, the critical stress for the transition from
viscous glide to climb should correspond to that for the breakaway of the dislocations from the
solute atmospheres,
W 2 c , (8)
c = m o 3
2 kT b
where Wm is the binding energy between the solute atom and the dislocation, co is the solute
concentration and is a factor dependent on the nature of the impurity cloud with a value
between 2 and 4. Due to the uncertainty in the values for the various parameters in this equation,
2

K.L. Murty and O. Kanert, J. Appl. Phys., Vol. 67 (1990) 2866.


O.D. Sherby and P.M. Burke, Progress in Mater. Sci., Vol. 13 (1967) 325.
4
J. Weertman, J. Appl. Phys., Vol. 18 (1957) 1185.
5
K.L. Murty, Scripta Met., Vol. 7 (1973) 899.
3

KL Murty

page 6

an exact transition stress (c) cannot be determined. However, the range for this critical stress
corresponds to 2x10-4E to 1x10-3E; E is the elastic modulus.17, 6 Such transitions were reported in
many class-A alloys.7 A review on the transition mechanisms in class-A alloys has been made
recently by Murty.8 The transitional stress in the present investigation does fall in this regime,
and thus we believe that the high stress region is due to the climb of edge dislocations. However,
the activation energy of 13 kCal/mole determined from both creep and ABI experiments is nearly
half that for self diffusion (QSD) in this alloy system. This value of Q agrees with the activation
energy for low temperature dislocation climb assisted by vacancy diffusion through dislocation
core. However, in that case, the stress exponent will be higher by a value of 2 or equal to 7.9, 15
It is possible that viscous glide is the rate controlling mechanism in both the regimes. At low
temperatures and/or high stresses, solute diffusion through dislocation pipes could dominate so
that the strain-rate would be proportional to the dislocation density,
= A1 e-Qpipe 3 . (9a)
since 2,
= A' e-QPipe 5. (9b)
This implies that at low stresses and high temperatures, viscous glide operates with the steady
state creep-rate varying as 3 with an activation energy close to that for solute diffusion through
the lattice. Our data is not sufficient to determine the activation energy for creep while cube-rate
dependence is observed. At high stresses, however, the stress exponent of 5 along with the
activation energy for creep of about one-half of that for lattice diffusion are in good agreement
with low-temperature viscous glide. Further work is needed to unequivocally prove this
contention.
Lead- Al loys
Pure lead, Pb5%Sn and Pb62%Sn alloy ingots, measuring about 0.5" thick x 1.5" wide x 5" long,
were obtained from Indium Corp from which appropriate specimens were made without any
further heat treatment. The optical metallographs depicting the grain structures of the three alloys
are included in figures 7a to 7c which were taken at 500X. It is clear that both pure lead exhibits
relatively large grain size while Pb5Sn depicts solid solution microstructure with relatively
smaller grain size. The lamellar eutectic microstructure is evident in Pb62Sn. Typical
microstructures at lower magnification are being taken for grain-size and/or phase size
evaluation.
6

P.Yavari and T.G. Langdon, Acta Met. Vol. 30 (1982) 2181.


T.G. Langdon, in Dislocations and Properties, The Institute of Metals, London (1985) 221.
8
K.L. Murty, in Proc. Seventh International Conference on Creep and Fracture of Engineering Materials and
Structures, J.C. earthman and F.A. Mohamed (Eds.) TMS (1997) 69.
7

KL Murty

page 7

a. Pure Lead (500X)

b. Pb5Sn (500X)

c. Pb62Sn (500X)
Figure 7. Optical Metallographs at 500X for pure Lead (top-a), Pb5%Sn (middle-b)
and Pb62%Sn Eutectic (bottom-c)
Strain-Rate vs Stress Results at Room Temperature:
ABI tests were performed on pure lead, lead with 5%Sn solid solution alloy and eutectic
Pb62%Sn at room temperature at varied strain-rates by varying the indenter velocity from 2x10-5
sec-1 to 2x10-2 sec-l. In many cases, relaxation during tests at the lowest strain-rate of 2x10-1 sec-1

KL Murty

page 8

precluded in obtaining reliable load-penetration curves. Strain-rates were evaluated at load


maximum using Francis' equation9 following the procedure outlined by Murty et al.10 True
stresses are evaluated at the maximum load along with true strain-rates. These results are first
analyzed assuming power-law dependence of the strain-rate,
= A n, where n is the stress exponent.
Figure 8 shows these data on log-log plot and the results for the three materials yielded the
following relationships,
= 4.8x10-17 12.4 for Pure Lead
= 1.3x10-26 15.7 for Pb5Sn
= 4.8x10-19 9.6 for Pb62Sn

Figure 8. Log-Log Plot of ABI Results at Room Temperature


Representing Power-Law Dependence
The relatively large values for n indicate that an exponential stress dependence may be
appropriate at these relatively high stresses (10 MPa is around 1.18x10-3 G. where G is the shear
modulus). Thus, these data are plotted on semi-log scale in Figure 9 which depicts lines with
varied slopes,
= 1.643x10-6 e0.62 for Pure Lead
= 2.94x10-9 e0.45 for Pb5Sn
= 4.50x10-7 e0.18 for Pb62Sn
9

H. Francis, Phenomenological Analysis of Plastic Spherical Indentation, Trans. ASME, July (1976) pp. 272-281.
K.L. Murty, F.M. Haggag and R.K. Mahidhara, Tensile, Creep and ABI Tests on Sn5%Sb Solder for Mechanical
Property Evaluation, Journal of Electronic Materials, in print.
10

KL Murty

page 9

These analyses indicate that at these stress levels at room temperature, strain-rates follow
deformation mechanism with stress-independent activation area as found earlier for SnSb.
Further data analyses will be performed following tests at elevated temperatures which are now
in progress. These data at room temperature do indicate plausible deviations at the lowest stress
level in pure lead which reveals trend towards power-law with far lower values for the stress
exponents which correspond to dislocation climb regime (of about 5 or so). The strengthening
due to Sn-solute additions is clearly noted in these data.

Figure 9. RT ABI Results Plotted on a Semi-Log Form


Representing Exponential Variation of Strain-Rate
Load Relaxation (using Cylindrical Indenter):
Load relaxation tests were performed using a 62mil diameter WC cylindrical indenter. Tests
were performed on SSM at constant indenter velocity (1x10-3 in/s) and load versus depth were
monitored to about 15 mils (approximately 22-25% strain) at which point the indenter was
arrested/stopped. The load relaxation is monitored for about 30 minutes from which the load-rate
versus load was followed; these correspond to the strain-rate versus applied stress once the
modulus and indenter cross-sectional area were taken into account. These tests were performed
on the three Pb-alloys at room temperature and 50C.
Figures 10a to 10c show the load relaxation data for pure lead, Pb5Sn solid solution alloy and
Pb62Sn eutectic alloy at room temperature, and clearly demonstrate the decreased load as a

KL Murty

page 10

function of time in a very orderly manner. Corresponding load-rate versus load are shown in
double-log plots for the three cases in figure 11 for pure lead, Pb5Sn and eutectic Pb62Sn.

Figure 10
Load vs time for Pb (a-top), Pb5Sn (lo-left) and Pb62Sn (c-right)

KL Murty

page 11

Figure 11
Load-Rate vs Load for RT for Pb, Pb5Sn and Pb62Sn

Figure 12
Load vs time curves for Pb (a-top), Pb5Sn (b-left) and Pb62Sn (c-right)

KL Murty

page 12

Similar study was made at 50C and figures 12a to 12c include load versus time curves
for the three alloys which are very similar to but shifted on the stress axis to lower stresses (due
to higher temperature). Figure 13 compiles and compares the load-rate versus load for the three
materials at 50C tests.

Figure 13
Double-Log Plot of Load-Rate vs Load
(Pb, Pb5Sn and Pb62Sn at 50C)
Varying slopes of these curves as a function of load (or stress) clearly indicate the
transitions in deformation mechanisms which include dislocation controlled creep at high
stresses and point-defect dominated mechanisms such as Nabarro Herring and/or Coble creep
mechanisms. In addition, in-situ recrystallization of lead alloys is known to give rise to
oscillatory behaviors as seen here.
As pointed out earlier, further data analyses will shed light on the underlying deformation
mechanisms but warrant more funding and time. These results indeed point out to the viability of
SSM in characterization of deformation kinetics of materials. It is recommended that these
results be extended to other temperatures and creep tests at constant load/stress using closed-loop
control be included as further study.

KL Murty

page 13

APPENDIX C

Final Report

Sample Tensile and ABI Test Results on


Aluminum 2024-T4 in the As-Received Condition

November 1998

10

11

12

13

14

S-ar putea să vă placă și