Site occupation and mechanical properties of iron-doped NiAl


Material Information

Site occupation and mechanical properties of iron-doped NiAl
Physical Description:
Duncan, Andrew John, 1965-
Publication Date:

Record Information

Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
oclc - 34428459
System ID:

Full Text







To Mom and Dad


Many people have been extremely helpful to me during this part of my studies and

deserve to be honored in recognition of their guidance, financial support and

encouragement. Dr. M. J. Kaufman directed me during my studies and deserves much

recognition for his efforts. The other members of my supervisory committee, Drs. R.

Abbaschian, R. Connell, F. Ebrahimi and A. V. Kumar, have been exceedingly

understanding and flexible. The guidance and financial support of C. T. Liu at Oak Ridge

National Laboratory can not go without mention and he was instrumental in the success of

this research. In addition, I. M. Anderson, M. K. Miller, E. D. Specht and J. Bentley

assisted in achieving successful site occupancy measurements on the materials in this

study. E. P. George, J. H. Schneibel, C. J. Sparks and others from ORNL were also

valuable in their council toward this project. R. Darolia and J. Dobbs, from G. E.

provided the single crystals examined in this study. R. D. Noebe and M. L. Weaver

contributed valuable technical input and compositional analysis. At the university, many

members of the staff and students have been helpful. To Thad Adams, Stan Bates, Andre

Costa de Silva, Vladamir Levit, Cindy Link and to all the others who have assisted in this

project, I would like to extend my most sincere gratitude.

I also wish to thank the people in my personal life that have encouraged me in this

undertaking. Cecil Carmichael, Alex and Sunday Cozzi, Dave Cheney, George and Lisa

Demmy, Dave and Jennifer Moore, Chris O'Gara, and Mark Weaver have been long

suffering, patient and kind during my hour of need. I would like to thank Tammy Myers

and the Lyle family for moral support, many hot meals and very good company. My entire

family has been very supportive and understanding. Most importantly, I give thanks to

God, who loves me through His Son, in spite of knowing me well.


ACKNOWLEDGMENTS.......................................................................... iii

A B STRA CT ........................................ .................. .... ........................ .. vi

1.0 INTRODUCTION............................................................................... 1

1.1 B asis for Study................................. .... ......................................... 1
1.2 Objective and Approach .................................................................... 3

2.0 REVIEW OF THE LITERATURE............................................................ 4

2.1 NiAl: Physical Properties .................................................................. 4
2.2 NiAI: Mechanical Properties............................................................... 6
2.2.1 Slip Systems ..................................... ................. ............... 6
2.2.2 Y ield Stress ..................... ....................... .... ........................... 7
2.2.3 Single C rystals ....................................................................... 7
2.2.4 Polycrystals................................ ........................................ 8
2 .2 .5 F racture............................................................................... 10
2.2.6 Elevated Temperature Deformation ............................................... 11
Brittle to ductile transition temperature.............................................. 11
C reep .................................................................................. .. 12
2.2.7 Effects of Stoichiometry and Purity............................................... 12
Stoichiom etry........................................................................... 12
Im purities .................................. .................... ..................... 13
2.3 Solid Solution Strengthening............................................................. 15
2.4 Alloying Effects ............................................................................ 17
2.4.1 General Behavior.................................................................... 17
Physical properties of iron-modified NiAl.......................................... 19
2.4.2 Mechanical Behavior................................................................22
Mechanical properties of iron-modified NiAl....................................... 25
2.5 Site Occupancy Determination of Substitutional Additions in Ordered
C om pounds ............................................. ... ........ ......................... 26
2.5.1 Theoretical Predictions ............................................................. 26
2.5.2 Experimental Determination of Site Occupancy ................................. 32
X-ray diffraction / density measurements........................................... 34
Atom Probe Field Ion Microscopy................................................... 34
ALCHEMI ................................. ................ ............................ 35

3.0 METHODS...................................................................................... 39

3.1 Alloy Production ........................................................................... 39
3.2 Thermal and Thermo-Mechanical Processing........................................... 39
3.3 Lattice Parameter Determination and Density Measurements ......................... 41
3.4 Atom Probe Field Ion Microscopy ...................................................... 43
3.5 ALCHEMI .................................................................................. 46

3.6 Mechanical Testing......................................................................... 50
3.6.1 B end T esting......................................................................... 50
3.6.2 M icrohardness....................................................................... 51
3.7 Characterization ............................. ............. ....... .... ....................... 51

4.0 RESULTS AND DISCUSSION ............................................................. 52

4.1 Site O occupancy ........................................ ..... ..... ...... ..................... 52
4.1.1 X-ray Diffraction/Density Measurements ........................................ 52
4.1.2 A PFIM ........................................... .................................. 55
4.1.3 ALCHEMI ......................................................................... .. 67
4.1.4 Summary of Site Occupancy Studies............................................. 74
4.2 Mechanical Properties ..................................................................... 77
4.2.1 Slip Behavior in NiAl and NiAI + Fe Single Crystals.......................... 77
4.2.2 Ductility and Flow in NiAl + Fe Polycrystals ................................... 78
4.2.3 Fracture Behavior of Dilute NiAl + Fe Alloys................................... 88
4.2.4 M icrohardness....................................................................... 91
4.2.5 Solid Solution Hardening.......................................................... 91
4.2.6 Summary of Mechanical Properties ............................................. 112

5.0 SU M M A R Y ................................................................................... 114


A X-RAY DIFFRACTION DATA........................................................ 116
B ALCHEMIDATA........................................................................ 117
C DISLOCATION ANALYSIS........................................................... 118

LIST OF REFERENCES ........................................................................ 122

BIOGRAPHICAL SKETCH.................................................................... 128


Abstract of Dissertation Presented to the Graduate School
of the University of Florida in Partial Fulfillment of the
Requirements for the Degree of Doctor of Philosophy



Andrew John Duncan

December, 1995

Chairman: Dr. Michael J. Kaufman
Major Department: Materials Science and Engineering

Aluminide intermetallics, such as NiAl, are important candidates for real life

applications because they are considered to be one of the few alternatives to conventional

alloys that offer any macroscopic ductility at ambient temperatures combined with

environmental resistance at high temperatures. A study to understand the influence of

solute atoms on the mechanical properties of the intermetallic compound NiAl has been

conducted using iron as the solute addition while varying the Ni/Al ratio. A matrix of

alloys containing 0.25, 2, 5 and 10 percent iron and different substitution schemes has

been selected 1) to examine the role of deviations from stoichiometry on the site preference

of iron in the B2 superlattice and 2) to understand the effect of iron on the ductility and

solid solution strengthening of NiAI at ambient temperatures. Site preference of iron is

determined as a function of deviations from stoichiometry and concentration using three

independent techniques: X-ray diffraction (compared with Archimedes density

measurements), atom probe field ion microscopy (APFIM) and a quantitative formulation

of atom location by channeling-enhanced microanalysis (ALCHEMI). The role of iron's

site preference in the generation of constitutional defects is discussed and estimates of

defect concentration are calculated from measured site preferences. The iron site

occupancies measured from APFIM do not completely agree with the results of the other

techniques; this is the result of a systematic error inherent to certain nickel deficient NiAl +

Fe alloys. The other techniques provide reasonable site occupancy data for iron in NiAl

and also agree qualitatively.

The mechanical properties of NiAl + Fe alloys at ambient temperature using three-

point bend testing are presented. Unlike reports in single crystals, iron is shown to have

no beneficial effect on the polycrystalline ductility of NiAl. Microhardness measurements

have been used to examine the effect of site occupancy on solid solution strengthening.

When iron is substituted for aluminum, the hardness is proportional to the square root of

iron concentration (4Ic) and obeys classical solid solution strengthening theories for metals.

When iron is substituted for Ni or both Ni and Al (Ni/Al = 1), the hardening response

deviates from this 4c relationship and a very pronounced hardening rate is observed when

iron is substituted for nickel. Finally, the correlation of the observed properties with the

role of site substitution, constitutional defects and overall iron concentration is addressed.


1.1 Basis for Study

Intermetallics have been the focus of considerable research for the past two decades in

the area of high temperature materials. The increase in the operating temperatures of gas

turbine engines towards the melting point of current superalloys has created a need for

new, revolutionary materials. These materials must exhibit good oxidation resistance,

favorable densities, sufficient strengths and high toughnesses at both ambient and elevated

temperatures. NiAl is of particular interest for these applications because of its combination

of good oxidation resistance, high thermal conductivity and low density. Unfortunately,

low strength in the unalloyed condition and low fracture toughness and plasticity at ambient

temperatures have limited the use of NiAl to non-structural applications (e.g., oxidation

resistant coatings).

Some property improvements have been achieved by alloying with potent solid solution

strengtheners (e.g., Ta, Hf, and Ti) to increase NiAl's low and high temperature strength

[1, 2]. However, Cotton et al. noted that ternary alloying additions did not behave

consistently with classical solid solution strengthening mechanisms [3]. Specifically, the

rate at which substitutional solutes strengthen NiAl is an order of magnitude higher than

classically observed in other systems. It has also been noted that the solution strengthening

rate (drt/dc) of an element does not vary linearly with the lattice misfit caused by the solute


Improvements in low temperature ductility have also been achieved by microalloying

with certain elements (e.g., Fe, Ga and Mo) [5]. Various mechanisms have been proposed

for this ductility enhancement, but no conclusive evidence has been provided in support of

these mechanisms. Hack et al. showed that thermal processing at intermediate temperatures

enhanced the toughness and ductility of NiAl at room temperature [6]. Weaver et al.

showed that carbon can cause dynamic strain aging in NiAl which may be counteracted by

the addition of reactive elements (e.g., titanium) [7]. It should be emphasized that

improvements in one area (e.g., strength) almost always detrimentally effect the other

(e.g., ductility). Furthermore, as significant as these advances appear, the lack of a

fundamental understanding of the mechanisms governing solid solution strengthening and

ductility limit significant further improvements in both areas.

Unfortunately, studying the fundamental mechanisms controlling the deformation

behavior of NiAl is complicated because of the compounding errors associated with its

processing and composition. Careful review of the literature indicates a broad scatter in

both yield strength and ductility values for "equivalent" compositions. This is primarily

because the mechanical properties of NiAl are extremely sensitive to stoichiometry, purity

and thermal history [6-8]. Hence, it is difficult to compare the subtle effects of

microalloying additions on properties when the segregation behavior of interstitials is not

well understood and Ni/Al ratio is not controlled. In addition, alloying with ternary

additions gives rise to constitutional defects in NiAl; these can cause dramatic strengthening

depending on how the elements are substituted. The site occupancy of these elements can

fluctuate depending on Ni/Al ratio, solute level and temperature. In order to fully

understand the behavior of NiAl alloys, a comprehensive understanding of site occupancy

and interstitial segregation and their effect on dislocation motion must be established.

1.2 Objective and Approach

The purpose of the present study was to determine the site occupancy of iron in NiAI as

a function of solute level and substitution scheme in order to understand the solution

strengthening and mechanical properties of iron-modified NiAl. The approach taken was to

examine a matrix of NiAl alloys containing up to 10% iron and various Ni/Al ratios. The

determination of site occupancies and constitutional defect concentrations estimated by

assuming conservation of atoms is also performed in order to observe general trends for

site substitution. This knowledge will then be applied to the mechanical properties

exhibited by the NiAlFe alloys.


2.1 NiAl: Physical Properties

NiAI has the B2 (CsCl) structure shown in Figure 2.1 [9]. This structure is simple

cubic, but is similar to a body-centered cubic lattice with Al atoms on the comers of the unit

cell and Ni atoms in the body-centered position. NiAl is a strongly ordered compound and

its long-range order parameter (LRO) has been measured to approach unity [10].

Furthermore, this compound is stable over a wide temperature range. The heat of

formation AHf of NiAl is negative and has been calculated to be somewhere between -63

and -72 kJ/mol [11]. This suggests that its bonding is very directional, being much

stronger between unlike atoms than like atoms. Low-angle structure factor calculations and

calculations from first principles both suggest the presence of extensive charge build-up

and transfer between unlike, nearest neighbors, leading to a mixture of covalent and

metallic bonding [12, 13]. As a consequence, NiAl is elastically anisotropic, i.e., its

stiffness coefficients (i.e., C11, C12, C44) differ depending on crystalographic orientation.

The Zener anisotropy factor (A= 2C44/[C I-C 12]) for NiAl has been measured to be 3.28


NiAl is a compound that, although stoichiometric at Ni-50 at. % Al, is stable over a

wide composition range (Figure 2.2 [13]). At 50 at. % Al, NiAl melts congruently at 1911

K (1638C). The melting point decreases with increasing deviation from stoichiometry and

peritectic reactions occur on either side of this compound on the phase diagram. On the Ni-

rich side, the peritectic reaction NiAl + L -> Ni3Al occurs at 1668 K (1395C) and the

Figure 2.1.





1 1000


Unit cell of the B2 crystal structure with <100>, <110> and <111>
directions indicated.

Weight Percent Nickel

0 t0 20 30 40 so0 0
Al Atomic Percent Nickel

Figure 2.2. Currently accepted phase diagram for the Ni-Al system [13].

solubility limit of Ni in NiAl is 69.2 at. % Ni (30.8 at. % Al). On the Al-rich side, a

peritectic reaction NiAl + L -> Ni2A13 occurs at 1406 K (1133C) where the maximum

solubility of Al is 58 at. %.

As in most intermetallic compounds with considerable solubility ranges, deviations

from stoichiometry result in the formation of constitutional defect structures in the lattice.

Specifically, Al-rich NiAl maintains an overall B2 structure by forming vacancies on the Ni

sublattice (VNi), while Ni-rich NiAl forms anti-site defects, i.e. Ni atoms on Al sites

(NiAl). Figure 2.3 shows how the lattice parameter and density change as a function of

stoichiometry [14]. As can be seen, the rapid drop in density and lattice parameter show

the impact of VNi'S on these properties. These defect structures drastically affect the

mechanical properties of NiAI [8]. According to the triple-defect model [15], thermal

defects in NiAl consist of one NiAl defect accompanied by two VNi'S on the Ni-sublattice

in order to maintain stoichiometry. The enthalpy of formation, AHt, of a triple defect in

NiAl is 183 kJ/mol which is large compared to that of FeAl (101 kJ/mol) where triple

defect concentrations are high [15].

2.2 NiAl: Mechanical Properties

2.2.1 Slip Systems

At low temperatures, NiAl deforms predominantly by <100>{ 110} slip [16]; this

supplies only three independent slip systems [17] and does not satisfy the Von Mises

criterion in which at least five independent deformation mechanisms are required for

achieving ductility in random polycrystals [18]. At present, reported room temperature

tensile ductilities vary between 0 and 2% elongation for polycrystalline, near-stoichiometric

NiAl [19-21 ] consistent with predictions based on the von Mises criterion. At higher

temperatures the ductility increases suggesting a thermally activated mechanism for

overcoming non-uniform deformation.

Similar to BCC metals and alloys, if <111> type slip were active in NiAl, the von

Mises criterion would be satisfied. However, such dislocations produce local disorder in

the B2 lattice. Specifically, a dislocation with b = a/2< 11> results in like atoms being

nearest neighbors, i.e., regions that are out of phase. The boundary separating the slipped

and unslipped regions is called an anti-phase boundary (APB) the energy of which depends

on the relative energies of A-A, B-B and A-B bonds. Consequently, for <111>

dislocations, a second dislocation with the same slip vector must glide along the slip plane

in order to annihilate this APB. Thus, most B2 intermetallics that exhibit <111> slip

contain pairs of dislocations separated by an APB; these are known as superdislocations.

Since NiAl has an extremely high ordering/APB energy [22], pairs of a/2

dislocations, or superdislocations, have yet to be observed. In fact, the critical resolved

shear stress (CRSS) for <111> slip on {112} planes has been measured to be -5.5 times

greater than <010> slip on {001 } planes at room temperature [23].

2.2.2 Yield Stress

Polycrystalline NiAl, as well as single crystals oriented for <100> slip, exhibit flow

stresses that increase rapidly with decreasing temperature below = 400 K. This is believed

to result from a high Peierls stress. Figure 2.4 plots yield stress vs. temperature and

clearly shows that at low temperatures the yield stress is heavily dependent on temperature

[14]. This is similar to the low temperature behavior of body centered cubic metals (e.g.,

vanadium). As temperature increases, some compositions show a plateau in the yield

stress vs. temperature relationship before dropping again at higher temperatures.

2.2.3 Single Crystals

In orientations that do not favor cube slip (i.e. <100>), NiAl single crystals exhibit

high compressive yield strengths (e.g., CYS-1350 MPa) at room temperature. Fraser et

al. showed that [001]-oriented crystals of NiAl are geometrically unstable, i.e.,

compressive loading usually results in localized kinking upon yielding [24]. Tensile

loading, on the other hand, leads to fracture before any macroscopic ductility is achieved

[23, 25, 26]. This behavior occurs because there is no resolved shear stress (ass) on any

of the preferred <100>{0kl } slip systems and, therefore, fracture occurs before slip is


Crystals oriented favorable for <100> slip have drastically lower yield points

(CRSS-85 MPa) compared with the [100]-oriented crystals [25]. Although substantial

deformation is possible in compression [27], only about 0-1% tensile elongation occurs

before failure in tension [25]. Recently, careful control of heat treatment and purity has

produced ductilities in excess of 5% [28]. This suggests that NiAl may, in fact, be ductile

at room temperature in single crystal form when impurities or processing variables are

controlled adequately.

2.2.4 Polycrystals

As mentioned before, failure to meet the von Mises criterion inhibits the room

temperature ductility of NiAl polycrystals. The effect of grain size on the ductility in NiAl

has been modeled and the existence of a critical grain size, below which NiAl would exhibit

ductile behavior, was suggested by Schulson et al. [29, 30, 31] and Chan [32]. These

models are based on the premise that, upon yielding, stresses build up at the grain

boundaries and microcracks form. If these flaws are small, they will undergo stable crack

growth during deformation until they become critical and catastrophic failure occurs [29,

30, 31]. The size of these microcracks is directly proportional to the grain size, i.e., the

stress concentrations (size of dislocation pile-ups) at grain boundaries are smaller when the

grain size is reduced.

Noebe et al. [33] used the J-integral approach to correlate values for ductility as a

function of grain size at temperatures above and below the brittle to ductile transition


Ni content (at.%)

Lattice parameter and density of NiAl as a function of composition [14].

2000 ooo I 1 1 1
1 Bowman et al. 1992 Pas
1800 [
Grain size = 10 pm
1600 Ni-50.6AI

1400Q Grain size = 30 pm
0 A Ni-50.6AI

i 1000 -

o 800 -

600 -

400 -

200 -
(b) | | ^ a^^ ]^
0 200 400 600 800 1000
Temperature, T (K)

1200 1400 1600

Yield stress vs. temperature for several polycrystalline alloys [14].

.289 I-

.288 [-

.287 -

.286 L-

Figure 2.3.

0 Taylor and Doyle 1972
o 0 Georgopoulos and Cohen 1977
V V Bradley and Taylor 1937
A Guseva 1951
0 Cooper 1963 5

- 4

Lattice parameter

-i 3


Figure 2.4.

temperature (BDTT), i.e., 700 K and 300 K, respectively. At 700 K, the model is in

agreement with the results by Schulson et al. and a critical grain size between 10 and 20 gim

is predicted. At 300 K however, the model predicts a maximum ductility of only -5% for

grain sizes as small as 0.1 gim. When it is considered that the smallest grain size reported

to date in NiAl is only 1 jtm [34], the limitations of this approach for improving ductility at

low temperatures become apparent. In addition, since reductions in grain size are known to

reduce the creep strength, such approaches are undesirable from the standpoint of high

temperature applications.

2.2.5 Fracture

NiAl fails in a brittle manner at temperatures even above its BDTTI (-550-700 K) [35].

Single crystals fracture primarily by cleavage. The preferred cleavage plane in NiAl has

been identified as { 110} [37]. Chang et al. [38] identified initial regions of {511 } cleavage

planes in single crystal bend specimens when { 100} cleavage was geometrically favored

until the crack could orient itself for propagation along the preferred { 110} planes. The

actual mechanism by which this occurs has not been explained.

The fracture mode in small-grain polycrystals tends to be intergranular. One reason is

that the insufficient number of active deformation mechanisms inhibits shear strain transfer

from grain to grain and stresses build up at boundaries causing premature failure at low

levels of plastic strain [29-33]. It has also been suggested by George etal. that NiAl has

intrinsically weak grain boundaries which limit ductility [20, 36]. Although grain

boundaries can be strengthened with boron additions, the cleavage strength in NiAl is

observed to be low. Hence, suppression of intergranular fracture does not significantly

increase ductility in polycrystals but merely caused cleavage fracture instead.

The fracture toughness of NiAl is also rather low at room temperature. As previously

mentioned, single crystals of NiAl have a KC ranging from 4 to 8 MPav/m depending on

orientation. Hack et al. [6] noted that the fracture toughnesses (KQ) for single crystals

varied substantially with heat treatment. Specifically, air cooling from 1573 K resulted in a

KQ of 16 MPav/m but reannealing at 473 K resulted in a KQ of only 3 MPaVm. The

fracture toughness of polycrystalline NiAl is not higher than single crystals; its value is in

the range of 4-7 MPavm [39, 40]. This suggests that grain refinement is not an effective

toughening mechanism in NiAl.

2.2.6 Elevated Temperature Deformation

Brittle to ductile transition temperature

The deformation characteristics of NiAI change drastically with increasing temperature.

As illustrated by Noebe et al. [33], as the temperature is raised, the yield point slowly

decreases and the fracture strength remains relatively constant until a critical temperature of

approximately 550 K is reached. Above this temperature, the fracture stress increases

considerably whereas the yield stress continues to drop. This is marked by a dramatic

increase in tensile ductility. The BDTT in NiAl has been attributed to the onset of <100>

dislocation climb [33] which, when accompanied with <100> glide, supplies the five

independent slip systems required for polycrystalline deformation [41]. Noebe et al.

observed an increase in the BDTT with increasing strain rate in polycrystalline NiAI, which

supported the conclusion that the BDTT was caused by a time dependent or thermally

activated mechanism (i.e. climb rather than glide) [35]. Field et al. argued that a diffusional

mechanism, such as climb of <100> dislocations, could not significantly contribute to the

overall deformation of NiAl at temperatures as low as 550 K [42] and proposed an

alternative explanation which involved the activation of additional slip systems above the

BDTT. This was supported by Lahrman et al., who reported that strain rate had little

significant effect on the BDTT of NiAl single crystals, suggesting that diffusion does not

limit deformation [24], and Zaluzec et al., who found <110> dislocations in compressed

<100> single crystals, implying that softening was due to glide of <110> dislocations [43].

Definitive proof of either mechanism has not been confirmed experimentally.


At even higher temperatures, dislocation climb becomes the dominant deformation

mechanism. Pascoe and Newey [44] characterized the deformation behavior of NiAl over a

large temperature range and observed a drastic reduction in the yield stress above room

temperature for stoichiometric NiAl. In addition, it has been shown that stoichiometric

NiAl has poor high temperature creep strength in comparison with current superalloys [1].

However, significant increases in high temperature strength can be achieved by alloying.

Solid solution, dispersion and precipitation strengthening have all been used successfully to

improve the high temperature strength of NiAI [1, 45].

2.2.7 Effects of Stoichiometry and Purity

Current processing techniques allow for careful control of composition, but a wide

range of properties are still reported in the literature for stoichiometric NiAl. This is largely

the result of the acute sensitivity of NiAl to point defects. Additionally, the presence of

impurities can cause rapid solution strengthening, strain aging or precipitation reactions

[46, 47]. Deviations from stoichiometry and impurites can drastically change the properties

of NiAl and, when not considered, may lead to misconceptions about mechanical behavior.

Although a large body of literature is available on this topic, only a brief summary of

related work will be discussed here.


As the composition of NiAl deviates from stoichiometry, pronounced hardening is

observed (Figure 2.5). Furthermore, no room temperature ductility is observed. Vedula et

al. [19] showed that less than 1% deviation from stoichiometry resulted in brittle behavior

at room temperature as well as an increase in the BDTT for extruded polycrystalline NiAl.

As noted above, deviations toward the Al-rich side of stoichiometry produce the most

potent hardening due to the presence of constitutional vacancies in the B2 lattice.

Specifically, constitutional vacancies have been reported to harden at a rate of 17.5 GPa/Vc

[48]. This hardening is magnified even more when it is considered that 1 % excess

aluminum will give rise to 2 % constitutional vacancies. On the nickel-rich side, anti-site

defects (Ni on Al sites) are produced and harden the lattice to a lesser extent (5 GPa/Vc).

Figure 2.5 illustrates the effects of stoichiometry on hardness.


Interstitial impurities also have a potent hardening effect in NiAl. A comparison of

powder-processed and conventionally-cast ingot materials in Table 2.1 shows that, in

general, the powder-processed alloys exhibit low (if any) ductility and seldom yield in

tension prior to failure [20, 21, 27, 49, 50]. Other work which compares high purity NiAl

with conventional purity NiAl suggests that impurities play an important role in low

temperature deformation [51, 52]. Cotton [51] noted a drop in the hardness of

polycrystalline NiAl as the purity of the starting materials was increased, while Weaver et

al. [52] noted the CRSS of high-purity single crystals was significantly lower than that of

conventional purity single crystals. In another study, Hack etal. [6] showed that annealing

at high temperatures and air cooling, designed to prevent segregation of impurities to

dislocations in NiAl, increased the room temperature fracture toughness and tensile

elongation to failure of NiAl single crystals to 16 MpaVm and 8%, respectively. In

addition, these authors observed a degradation in properties if the material was re-annealed

at lower temperatures (473 K). Strain aging was proposed as a possible mechanism for

this change in properties. Weaver also showed that alloys doped with carbon underwent

dynamic strain aging at = 600 K [54].

In other studies of interstitial effects on NiAl, carbon and boron were determined to

increase the yield strength at rates of over 1700 and 3870 MPa/at. %, respectively [20, 36].






45 50 55 60

Atomic Percent Ni

The effect of deviations from stoichiometry (Ni/Al = 1) on the hardness of
NiAl [53].

Figure 2.5.

Later, MC and MB2 precipitation was observed in these alloys using atom probe field ion

microscopy (APFIM) [47]. Similar precipitates were also observed in NiAl single crystals

[8]. The metallic components of these carbide/boride precipitates were determined to be

trace elemental impurities of Ti, V and Cr. It was suggested that a significant portion of

these high hardening rates was a result of precipitation hardening.

Table 2.1. Room temperature mechanical behavior of polycrystalline NiAl
and its alloys in tension.

Comp. Grain Size Process Yield UTS Plastic
(at.%) (RIm) Method (MPa) (MPa) Strain Reference Year
_ _ _ _ (%) _
Ni-50A1 -40 PM, HP 740 0 [55] 1952
Ni-43A1 no data cast 152 0 [56] 1954
Ni-50A1 -200 cast 104 0 [57] 1957
Ni-50.5A1 no data cast, extr. 179 207 2-4 [21] 1966
Ni-50.3A1 10-16 cast, extr. 235 324 2.5 [50] 1989
Ni-50.6A1 33 PM, extr. 284 0 [49] 1989
Ni-50AI 5 PM, extr. 220 401 0-1 [33] 1990
Ni-50A1 30 115 220 2.5 [58] 1990
Ni-50AI -30 cast, extr. 154 230 2 [20] 1990
NiAl- -30 cast, extr. 329 0 [20] 1990
NiAl- -30 cast, extr. 200 246 2.0 [36] 1991
NiAl- -30 cast, extr. 332 332 0.7 [36] 1991
NiAl- -30 cast, extr. 336 0 [20] 1990

2.3 Solid Solution Strengthening

Solid-solution strengthening in metals is known to result from several contributing

mechanisms. The alloying additions are frequently represented by a random array of point

defects that provide a restraining force on dislocations. The effect increases the flow stress

(to) in an additive manner, i.e., t = to + ts where ts is due to solid solution hardening.

Fleischer [59] noted that the flow stress can be estimated by the stress required to overcome

a restraining force, Fmax, on a dislocation per unit length, L, divided by the burger's

vector, b.

t = Fmax/b L

Depending on dislocation flexibility, L = b / V/2c for flexible dislocations or L = b / VcO for

partially flexible dislocations, where 0 is the bending angle of a dislocation around an

obstacle and c is the solute level in mole fraction.

The mechanisms contributing to the solid solution strengthening are elastic strain

interactions, modulus interactions, stacking fault, electrical and chemical interactions and

both short and long-range order interactions. Most of these interactions are system specific

and few have actually been incorporated into models for solid solution strengthening. By

and large, the dominant interaction is the elastic interaction associated with strain in the

lattice. The misfit strain, ea, is the normalized change in lattice parameter, ao, as a function

of composition (-a = 1/ao. da/dc). Mott and Nabarro related strengthening to the square of

the misfit strain:

t = G 2 c

where T is the flow stress and G is the shear modulus. However, it has been observed that

defect solid solution strengthening in NiAl does not follow this relationship [4].

Modulus interactions were considered by Fleischer and a strain misfit parameter, Es,

was identified as a combination of strain and modulus contributions, i.e., Es = EG ax a,

where a is an adjustable parameter applied to fit the data of a specific system. An

expression for the shear stress was determined based on the equation:

t = (G 1.5v/c) / 700

Asymmetrical lattice distortions are caused by defects that strain the lattice in one

direction preferentially and tend to harden metals more rapidly than atoms that cause

spherical distortions. A classic example of a system that exhibits tetragonal distortions is

carbon in iron where a hardening rate of (G/20)Vc is observed due to carbon atoms

occupying octahedral sites such that a distortion in <100> directions occur in relation to

other directions. Tetragonal distortions are known to strengthen much more rapidly

because of their interactions with screw dislocations, which are not affected by spherical


Electrical interactions between solute atoms and dislocations are perceived to arise from

the permanent charges carried on both. Solute atoms that have a higher valence state than

the solvent are thought to carry a net positive charge, which is believed to react with the

electron charge dipole present at the dislocation core. No experimental proof has been

obtained for this phenomenon in metals. Chemical interactions come as a result of stacking

faults or anti-phase boundaries (APB's) that are formed as a result of partial or super-partial

dislocations passing through the lattice creating a faulted region of energy Ya. Alloying

additions that are known to change the stacking fault energy and anti-phase boundary

energies in metals should also be important in intermetallics, i.e., they can cause pinning of

dislocations and, therefore, increase the shear stress required for slip by Aya / b.

2.4 Alloying Effects

2.4.1 General Behavior

Much of the literature on substitutional alloying additions has been summarized by

Cotton et al. who grouped the elements into three different classes (e.g., Types A, B and

C) according to the ternary NiAl-X phase equilibria (Figure 2.6) [3]. Type A elements

form ternary intermetallic phase(s) with NiAl and usually have relatively low solubility.

These elements have been investigated in detail in an effort to improve the high temperature

strength of NiAl. Strengthening is achieved by a combination of solution hardening and

precipitation of intermetallic phases (e.g., Ni2AlX and NiAlX, where X =


[ AI-Ni-X ternary phases)
NiAli-X quasibinary eutectic

High solubility in NiAl and/or forms B2 aluminide

Figure 2.6. The phase equilibria of many NiA1-X systems illustrated on a portion of the
periodic table [3].

Ti, Zr, Hf, Nb and Ta) in many of these ternary systems [1, 2], where the solute elements
exhibit a strong trend to occupy the Al site in NiAl. Type B elements form NiAl + a

quasibinary eutectics and some systems (e.g., NiAl-Cr) are viewed as promising

candidates for precipitation strengthening because of the large changes in solubility with

temperature. Type C elements exhibit extended solubility in NiAl and, in some cases, solid

solution (Ni, X)A1 regions are present in the phase diagrams.

Physical properties of iron-modified NiAl

NiAlFe alloys have been of interest since the mid-1930s for use in magnetic

applications and some characteristics of this system make it of interest today. A region of

extended solubility exists between FeAl and NiAI at high temperatures. Lipson and Taylor

[60] constructed a ternary isotherm of this (Ni, Fe)Al phase field and measured the lattice

parameter as a function of composition. They determined a steep decrease in lattice

parameter as alloys became aluminum-rich and the presence of constitutional vacancies was

noted. When iron was substituted directly for nickel, the lattice parameter remained

relatively constant, increasing slightly as the composition approached FeAl. Figure 2.7 a

and b are ternary Ni-Al-Fe isotherms at 1523 K and 1023 K [61]. The extended P phase

field between NiAl and FeAl is apparent at high temperature but at 1023 K the
encroachment of the a + P two-phase field can be seen. In Figure 2.8 [62], it is evident

that the miscibility gap impinges upon the middle of the P phase field but does not extend to

the aluminum-rich or nickel-rich parts of the phase field.

Stoichiometric NiAl has a low magnetic susceptibility. This changes, however, when

iron is added to the system. For binary NiAl and ternary NiAl + Fe with excess aluminum

atoms, Goldberg et al. [63] observed Pauli type paramagnetism. For all other NiAlFe

alloys, Curie type paramagnetism is observed. This means that additions of iron cause an

increase in magnetic susceptibility which is sensitive to temperature. When iron is added

for nickel, the magnetic susceptibility increases a small amount. This was rationalized by


20 80
MO 40/ +60
^/o.,. /.\
60// L 4
80 20
Fe 20 40 60 80 Al
Weight % Al



wt. % A]

Figure 2.7. Isothermal sections of the ternary phase diagram for the Ni-Al-Fe system
for 1523 K (a) and 1023 K (b) [61]

T a 1 a2


Fe FeAl At

Figure 2.8. Schematic representations of the miscibility gap in the Ni-Al-Fe system. A
a three dimensional image showing the evolution of this miscibility gap

noting that iron needs 4 electrons to fill its 3d shell while nickel only requires two. Since

aluminum can only supply 3 valence electrons, the magnetic susceptibility increases

slightly. On the other hand, when iron is substituted for both nickel and aluminum or

aluminum only, the magnetic susceptibility increases rapidly. This was explained by

noting that the iron atoms that occupy aluminum sites align their moments with the nickel

atoms on the other sublattice. In fact, Fu et al. [64] recently calculated the local magnetic

moments resulting from placing iron on nickel and aluminum sites to be 0. 1 IB and 2.4 gIB,

respectively. Lim et al. [65] also observed the magnetic susceptibility as a result of

precipitation of a-Fe in a Ni-44A1-12Fe alloy. Thus, a relatively large range of magnetic

properties exists in NiAl + Fe alloys depending on iron's site occupancy (substitution


2.4.2 Mechanical Behavior

While a vast amount of research has been devoted to alloy development in NiAl, the

effects of ternary alloying additions on the mechanical properties of NiAI are not well

understood. This is due largely to the complicated and interdependent influence that

stoichiometry, alloying additions and purity have on NiAl. Specifically, alloying additions

usually change the effective stoichiometry of NiAl because of their site preferences in the

lattice [51]; this in turn leads to property changes that depend on both the alloying addition

and the resultant constitutional defects.

Many investigators have attempted to promote <111> slip in NiAl by alloying with

elements from all three categories [51, 66-68]. Law and Blackburn reported <111>

dislocations in alloys that contained either chromium or manganese yet observed no

ductility increase. Later, Cotton et al. [51] determined that chromium did not, in fact,

produce <111> slip. Hong and Freeman [22] used first principle calculations to predict

that vanadium, as well as chromium and manganese, should increase the probability for

0.5 1 1.5 2
Ternary alloying addition (atomic percent)


0.3 0.6 0.9 1.2
Ternary alloying addition (atomic percent)


Mechanical behavior of NiAl single crystals alloyed with iron, gallium and
molybdenum a) ductility and b) flow stress.









Figure 2.9.

4 _-T

<111> slip by lowering the APB energy. These calculations were conducted using the

assumption that vanadium would substitute for aluminum. However, Darolia et al. [67]

observed no <111> dislocations or increases in ductility as a result of vanadium additions.

Consequently, the possibility that additional slip systems can be activated in NiAI as a

result of alloying at concentrations less than 10% solute appears to be negligible.

Enhancements in the ambient temperature tensile ductility have been reported in NiAl

single crystals as a result of small (microalloying) additions of iron, gallium or

molybdenum (see Figure 2.9) [1, 4]. In all cases, the ultimate tensile strength increased

with increasing ductility. Possible explanations for this effect have been suggested to

involve solute/vacancy, solute/interstitial or solute/dislocation interactions, although no

definitive evidence has been presented to date. Polycrystalline alloys containing iron and

gallium additions were studied and found to exhibit similar or lower ductility than binary

stoichiometric polycrystals after similar processing [69, 70].

Ternary additions to NiAl can produce complex defect structures and result in potent

strengthening effects when the ternary addition exhibits a strong site preference [3, 51 ].

For example, if an element, X, prefers the Ni-site and is added for aluminum, NiAl or XA1

defects result. Likewise, when an element that prefers the Al site is added for nickel, the

result is the production of vacancies on the Ni sublattice (VNi). Hence, solution

strengthening in NiAI is heavily dependent on substitution strategy and the site occupation

behavior of the ternary additions. Cotton et al. [3] surveyed a wide variety of ternary

alloying additions and found that strengthening rates were heavily dependent on the

substitution approach. Table 2.2 lists the solution strengthening rates for several alloys

where it is important to note that these strengthening rates tend to be linear with

composition (a = (da/dc) c). Most alloying additions resulted in high strengthening rates

(e.g., do/dc = G/7) although some (e.g., interstitials and refractory elements) exhibited

much higher rates (e.g., do/dc = 7G). Fleischer [59] studied solid solution strengthening

in metals and found that substitutional elements strengthen at rates of G/100 < do/dc <

G/10 whereas interstitial solutes have rates in the range 2G < da/dc < 9G.

Table 2.2 Solid solution strengthening of ternary additions in NiAl

Atomic Substitution Aa/Ac Aa/Ac
Solute Radius (A) Scheme (MPa/conc.) 1/conc. Reference
Ni 1.25 Al 10,800 G/5 [511
Vacancies = 1.0 Ni 21650 G/3 [51]
Ni 1.25 Al 17,000 G/3 [48]
Vacancies =1.0 Ni 5,100 G/30 [48]
Cu 1.28 Ni 10,800 G/7 [3]
Cu 1.28 both 11,500 G/6 [3]
Cu 1.28 Al 5,000 G/14 [3]
Fe 1.24 Al 23,000 G/3 [70]
Cr 1.25 both 28,500 G/3 [51]
Ga 1.22 Al 25,000 G/3 [70]
Be 1.12 both 10,000 G/7 [20]
Y 1.79 Al 200,000 3G [71]
Hf 1.59 Al 432,000 6G [71]
Zr 1.58 Al 470,000 7G [72]
B 0.46 both 427,500 6G [36]
C 0.74 both 193,300 3G [20]

Mechanical properties of iron-modified NiAl

Jung et al. [73] studied creep in (Ni, Fe)Al alloys and observed a maximum in flow

stress at 10 % iron. This effect was most dominant at 1023 K but showed increased

strength over a range of temperatures. A minimum in the interdiffusion coefficient was

also observed and suggested as an explanation for the strength anomaly in these alloys. As

mentioned above, Darolia et al. [5] observed an increase in tensile ductility with minor

additions of Fe, Ga and Mo in NiAl single crystals. In the case of iron, the improved

ductility was accompanied by a slight decrease in the yield strength, whereas in

polycrystals, other authors observed that dilute NiAl + Fe alloys exhibit higher yield

stresses, lower ductilities and higher BDTT's than stoichiometric NiAl [74, 69]. Hence,

the reported beneficial effect of iron on the ductility in NiAl is confined to single crystals.

2.5 Site Occupancy Determination of Substitutional Additions in Ordered Compounds

2.5.1 Theoretical Predictions

Site occupancy in ordered materials has been determined to play an important role in

understanding the mechanical behavior of ordered compounds and many methods have

been developed to predict site preference behavior. Theoretical predictions can be made

based on the enthalpies of formation, electron to atom ratio, relative size and solubility

lobes in ternary isotherms. In some cases, authors have made quantitative predictions of

site occupancies based on these considerations. Experimental verification of site

occupation requires an array of characterization techniques where a wide variety of

sampling size and precision is available.

The most commonly used method for predicting site preference in alloys is to apply the

Hume-Rothery rules for solutions. These four rules indicate that for an addition to have

high solubility, it must (1) be similar in size, (2) have the same crystal structure, (3) have

the same valence and (4) have similar electronegativity as the parent atom(s). In a

compound like NiAl, where Ni and Al are elements of different radii, valences and

electronegativities, an alloying addition would be expected to substitute preferentially for Ni

or Al based on its greater similarity to one over the other. Table 2.3 lists these factors for

several elements including Ni and Al; from these characteristics, site preferences of alloying

additions in NiAl are predicted. For example, iron is expected to occupy the nickel sites

because of its similar atomic radius, crystal structure, valence and electronegativity.

From a large study of different compounds, Hume-Rothery [75] noted that similar

compounds tended to exhibit similar electron to atom ratios (e/a). For example, the B2

compounds NiAl FeAl, and CoAl all have e/a = 3/2, i.e., 3 electrons from Al, 0 for Ni,

Fe, or Co and 2 atoms/unit cell. He also suggested that the stability, defect structure, and

site preference of substitutional elements in compounds are often determined by the e/a ratio

when relative size, electronegativity and crystal structure effects are not dominant. Nickel

contributes no valence electrons to the structure because the 3d shell accepts its valence

electrons from its 4s shell. Vacancies on the nickel sublattice result in Ni-deficient NiAl

because aluminum anti-sites increase the e/a ratio, while nickel anti-sites are stable in Al-

deficient NiAl because they reduce the e/a ratio in the structure.

Table 2.3. Physical constants of metals in relation to NiAl and expected site preference
elements based on the Hume Rothery Rules for solid solution.

Atomic Equilibrium Electro- # of electrons
Element radius crystal Valence negativity added to (e/a) Expected site
__ (A) structure Ni preference

Ni 1.243 FCC +2 1.8 0 Ni

Al 1.432 FCC +3 1.5 3 Al

Fe 1.241/1.269 BCC/FCC +2,+3 1.8 0,1 Ni

Cu 1.278 FCC +1 1.9 1 NiorAl

Co 1.253 HCP +2 1.7 0 Ni

Mo 1.363 BCC +4 1.8 4 Al

Cr 1.249 BCC +3 1.6 3 NiorAl

Ti 1.475 HCP/BCC +4 1.6 4 Al

Nb 1.426 BCC +4 1.7 4 Al

Zr 1.616 HCP +4 1.5 4 Al

Ta 1.43 BCC +5 1.3 5 Al

This methodology can be applied to alloying additions in the absence of other more

dominant factors. Table 2.3 shows the number of electrons that would be contributed to

the structure if the alloying addition were to occupy a nickel site. As is the case with

nickel, it is assumed that iron and cobalt fill their 3d shell before contributing valence

electrons to bond structures, therefore, they substitute readily on nickel sites. Copper, on

the other hand, has its 3d shell full and contributes its single valence electron to the bond

structure; in this case, copper would be expected to favor the aluminum site.

An empirical method for site occupancy prediction involves examining the solubility

lobes of a compound with a ternary alloying addition [76]. Figure 2.10 is a schematic

representation of the Ni3Al + X phase fields in ternary Ni-Al-X isotherms. If a marked

increase in solubility is observed when X is substituted for nickel, it is clear that X prefers

nickel sites. Since binary Ni3Al has a narrow solubility range, the solubility lobes of Ni3Al

+ X alloys exhibit elongated shapes more readily than the wide single phase region of

NiAl. However, this technique can be applied to some ternary systems, such as NiAl + Nb

and NiAl + V. As can be seen from the Ni-Al-Fe phase diagram in Figure 2.8, an

isomorphous region of extended solubility exists between NiAl and FeAl [61]. As

temperature drops, this solubility region becomes restricted in the middle (Ni/Al = 1) but

remains extended for constant nickel (Ni = 50%) and constant aluminum (Al = 50%)

substitution schemes. This suggests that iron may occupy either site in the B2 structure.

Quasi-chemical models considering nearest-neighbor bond energies (H) and assuming

random mixing of atoms on each site have been used to determine site preferences in

compounds. Ochai et al. used this technique to predict ternary site distributions in Ni3X

compounds [76]. Kosla [77] later adopted this method to predict the site preference of

chromium in FeAl. This method neglects entropy considerations and thus is independent

of temperature. In this model, interaction parameters VAB, VAC and VBC for a given

composition are calculated to determine the relative energies for each bonding

configuration. VAB is the energy difference between conditions of completely mixed and

unmixed and is calculated the relation:

V = [-P e (AD)2 + Qo (Anws1/3)2 R]

where P, Qo and R are constants for a particular class of bonding (i.e., transition metal

bonds or transition metal-non transition metal bonds). The difference in chemical potential

for the elements in the bond, A
cells, Anws, are provided by Mediema et al., while e is the elementary charge [78]. The

relative site preference of an alloying element C in compound AB (e.g., NiAl) is

determined by plotting the interaction parameter ratios VAC / VAB vs. VBC / VAB. Figure

2.11 is a plot of these parameters for numerous elements in NiAl. To estimate the relative

site preferences, construction lines with a slope of one (i.e., VAC / VAB = VBC / VAB + y-

intercept) will mark elements with similar potential to occupy the sublattice, B. As can be

seen, Ti, V, Zr, Mo, and Ta all have a strong affinity to occupy aluminum sites, while Fe

and Mn appear to exhibit a slight preference for aluminum sites. The noble elements (i.e.,

Pd, Pt, etc.) seem to prefer nickel sites.

Kao etal. [79] proposed a model in which both enthalpy and entropy were considered

for dilute solutions in B2 compounds. The site occupancy of a given alloy was determined

to be:

NaA NIPC / NI3A NOC = exp [2K/RT]

Quantitative site occupancies were calculated for a temperature of 1300 K. The

composition that was required to force all of element C onto the a (or Ni) sublattice was

given for several different dilute solutions. In Table 2.4, almost all elements selected

require a composition of > 50% aluminum suggesting that most elements prefer to occupy

aluminum sites. However, palladium and cobalt have a high affinity for aluminum and

favor nickel sites in order to have aluminum nearest neighbors. This study also suggested

that iron would partially occupy the aluminum sublattice until the stoichiometry was such

that vacancies were being produced on the nickel sublattice, at which point, it would favor

Ni sites.

50 40 30 ^S20 1 ;50Nl
--- C

Figure 2.10. Ternary isotherm of Ni3AI + X alloys showing site occupancy
determination by the solubility [76].





0.4 0.8 1.2 1.6 2


Figure 2.11.

Energy plot of interaction parameters VAC / VAB vs. VBC / VAB showing
the relative site preferences of element (C) on Ni (A) or Al (B) sites.

Fu et al. used first principle calculations to determine the site preference of iron in NiAl

and found that site occupancies varied with solute level, stoichiometry and temperature

[80]. When iron was substituted for nickel, the site occupancy of iron on the nickel

sublattice was predicted to be 100% and independent of temperature and concentration up

to 10%. However, when iron was substituted for aluminum it was predicted to favor the

aluminum sublattice at low temperatures with the fraction of iron on Ni sites increasing

with increasing temperature and decreasing iron concentration. Figure 2.12 shows the

predicted site occupancies from this study when iron is substituted for aluminum.

Table 2.4. Composition at which 100 %

of element C would occupy the Ni site at 1300 K

2.5.2 Experimental Determination of Site Occupancy

Numerous techniques have been used to determine the site preferences of ternary

additions in ordered compounds. Three techniques (X-ray diffraction, atom probe field ion

microscopy and ALCHEMI) were used in the present study to determine site occupancy

Dilute Addition Absolute Site Preference

Pd Ni

Co Niif > 42.8%Al

Cr Niif > 50.4 % Al

Cu Niif > 50.4 % Al

Fe Niif > 50.5 % Al

Mn Niif > 51.8 % Al

Ti Niif > 52.9 % Al

V Niif > 53.8 % Al


CL 80


o Ni-site
M-. --o-- AI-site
0) n--'0

2- 20 E
C) 20 El-

0 a---- I I

0 2 4 6 8 10
at. % Fe

Figure 2.12. Predictions for the site preference of iron as a function of concentration
calculated by first principles calculations after Fu et al. [80].

and will be discussed here. In addition, previous relevant studies on NiAlFe systems will

also be discussed.

X-ray diffraction / density measurements

Chang et al. [81] estimated vacancy concentrations in FeAl as a function of

stoichiometry and temperature by comparing the lattice parameter, measured using X-ray

diffraction, with the bulk density, measured using the Archimedes method. In the case of

NiAl, the concentration of thermal defects is relatively low, so site occupancies of ternary

additions can be determined if vacancies are produced as a result of the substitution


Comparing superlattice to fundamental peak ratios is another technique for determining

site occupancies in a binary compound when the elements involved have different

scattering factors. However, the situation becomes complicated when a ternary addition is

distributed on two sites that exhibit similar X-ray scattering factors [82, 77]. This is a

problem with the Ni-Al-Fe system because nickel and iron have very similar scattering

factors. One way to circumvent this problem, is to fine tune the energy of the incident X-

ray beam near the K absorption edge of one of the elements in order to minimize the

scattering by that element and maximize the difference in the scattering factors, thereby,

increasing the accuracy in the measurement technique [77]. This approach has been used to

determine site occupancies in Fe(Al, X) and Cu(Au,X) compounds with accuracies of

about 1% [83, 84]. Unfortunately, this is not an option for conventional X-ray diffraction

analysis equipment and requires a synchrotron radiation source where the

wavelength/energy can be varied to the appropriate values.

Atom Probe Field Ion Microscopy

The field ion microscope (FIM) is used to image the surface of solids at high

magnifications and utilizes the small radius of curvature of a needle-like specimen to

achieve an image at magnifications of at least 106 [85]. This technique uses high voltages

concentrated at a specimen tip to ionize gas particles and accelerate them towards a channel

plate where the image is produced. When this microscope is fitted with an energy

compensated mass spectrometer, bulk compositional information can be acquired by

evaporating atoms from the surface with a voltage pulse and catching them with the mass

spectrometer (i.e., atom probe field ion microscopy, APFIM). This technique has a rather

small sampling volume (typically about 10,000 unit cells) but little manipulation or

correction of the data is required.

APFIM has been shown to be very useful in the study grain boundary segregation,

spinodal decomposition and clustering [85-87]. Additionally, several studies on the site

occupancy of ternary additions in aluminide intermetallics have established the atom probe

as a reliable source for site occupancy determination. Miller et al. [83] determined that, in

Ni3A1, hafnium occupies Al sites preferentially while iron appeared to occupy either site.

Kosla et al. [83] used this technique to determine the site occupancy of Cr in FeAl and

determined that Cr occupied aluminum sites, which agreed with X-ray diffraction using

synchrotron scattering. Finally, Jayaram et al. [88] determined that, in NiAl, Be occupied

Al sites only.


Atom location by channeling-enhanced microanalysis (ALCHEMI) is a relatively simple

technique for extracting site substitution information. It requires a transmission electron

microscope fitted with an energy dispersive X-ray spectrometer. Since a probe size of

approximately 200 nm is used on approximately the same sample thickness, the sampling

volume is roughly 3 x 108 unit cells. The site occupancies of alloying elements is

determined by comparing the variation of characteristic X-ray intensities (Ii) of

constituents, i, as a function of the crystal lattice orientation, , relative to the incident

electron beam. These variations result from different thickness-averaged electron

intensities 2 j 12 on different sublattices in the crystal, which are caused by the formation

of a standing wave across the unit cell due to dynamical electron diffraction. When the

standing wave maxima coincide with a given sublattice of the crystal, X-ray intensities

from atoms occupying this sublattice increase. These effects are negligible at crystal

orientations where no low-order Bragg reflections are strongly excited Tj = TO for all sites

Originally, the site preference of an impurity element, k, on a site occupied by marker

element j was determined by comparing its X-ray intensity ratio of channeling to non-

channeling condition ((IjC/Ij0 i0/Iic)) to the ratio of the sublattice marker element (j).

Taft0 and Spence [89] first used this technique to study cation ordering in spinels. But for

light elements, characteristic X-ray intensities are not as drastically affected by channeling.

This is due to ionization delocalization, which arises because weakly bound inner shell

electrons can be ejected by incident electrons that are a substantial fraction of the interatomic

spacing from the site of the atom. The variation of characteristic X-ray intensity ratio of

delocalized elemental lines is smaller than that of completely localized elemental lines. In

studies of site occupancy in aluminide intermetallics, quantification of results was reported

to be difficult due to this phenomenon [90, 91]. Munroe et al. [92] found that

delocalization effects increased with increasing accelerating voltage when parallel to the

[100] zone. Walls [92] and Pennycook [93] developed techniques to correct for

delocalization by multiplying intensity ratios by some constant. Walls also accounted for

the presence of marker element anti-sites and their influence on the site occupancy values.

Rossouw et al. [94] considered the statistical error associated with the simple ratio method

(IC/Ii0) and found large errors when channeling effects were small. To minimize this,

multivariate statistical procedures were incorporated into the analysis. Instead of a simple

ratio method, the signal of the additional element was plotted as a linear combination of the

marker elements for various orientations. The coefficients of this linear function yield the

distribution of the additional element on the lattice sites. More recently, Anderson et al.

[95] have incorporated both delocalization and anti-site defect corrections into a statistical

formulation of ALCHEMI for a multiple sublattice system in which error propagation is


The following formulation is a statistical ALCHEMI method developed by Anderson et

al. [95]. A system with elements i, of mole fraction fi, on sites j, of site fraction nj is

selected for ALCHEMI analysis. The fraction of atom i on site j is Pij. From conservation

of sites and atoms it can be written:

Ii fi = 1 Ij nj = 1 Yj pij = 1

The correction for ionization delocalization for element i is treated as a linear parameter (Li)

which is characteristic of the atomic shell and independent of diffracting conditions. The
characteristic X-ray intensity Ii, for each element i, at crystal orientation , is assumed to

be a linear combination of the thickness-averaged electron intensities at each sub-lattice

(| 'j() 12, fraction Li) and those averaged over the unit cell ( 0() 12, fraction (1-Li))
Ii() = Ki fi {Li y pij I Vj(4)12 + (1-L,) |yo( 12 }

Ki is a constant for a selected area in the specimen and does vary among spectra. The

normalized intensity (r, = Ii()/Ii(0)), is the intensity at orientation () normalized to the

intensity at the non-channeling orientation (0). The r-value (rk) of a specific alloying

element (k) is plotted as a function of the n marker elements j.

( -l-)= I Xkj ( rj 1)

The correlation coefficient akj is related to Xkj by correcting for delocalization. Lkj is

independently determined by a least square fit of r-values from channeling data taken from

planes of equal composition over a range of orientations (Lkj = drk/drj ).

Xkj = LjLk akj = Ljk Xkj

If no anti-site (i on site j) or vacancy defects are present, this is the fractional site
occupancy of additional element k on sublattice j. The sum of Xkj over all sublattices j must

equal 1. If defects are present, Xkj must be adjusted to conserve stoichiometry. For anti-

site defects of fraction Pij, the fractional site occupancy is given by:

pkj = Y Xki Pij

In the case of iron in NiAl, two marker elements (i = Ni, Al) and iron (k = Fe) are

distributed on the two sublattices ('Ni' and 'Al'). The site occupancy of iron on the nickel

site (pFe'Ni') is:

PFe'Ni' = XFeAI PAI'Ni' + XFeNi PNi'Ni'

If we assume that there are no anti-sites (pij = 1 if i = j, Pij = 0 if i # j), the total fraction of
atoms on majority site j (fTj) is equal to fj + fk Xkj. Hence, the vacancy fraction fvi on

minority site i, can be calculated by the following:

fvi = (fTj-nj) / fTj(1-nj)

where nj is 0.5 for B2 crystal structures. For example, the case of aluminum-rich NiAl

(i.e., Al isj and Ni is i) with iron the vacancy concentration is given by:

fVNi = (fAl + fFeXFeAl 0.5) / (fAl + fFeXFeA1) (1 0.5)
The error analysis of this method is determined by the chi-square deviation from the

planar fit correlation coefficients akj. This is proportional to 1/V-i (N = number of Fe

atoms) and does not reflect systematic errors from the technique.


3.1 Alloy Production

The alloys for this study were produced by non-consumable arc melting on a water-

cooled copper hearth under a high-purity argon atmosphere. The metals used to produce

these alloys were of high purity (> 99.95% pure) i.e. Ni (99.95%), Al (99.99%) and Fe

(99.99%). The individual components of the alloy were consolidated and melted into

button form, turned over and re-melted a minimum of five times to insure homogeneity.

Weight loss was determined for each ingot to insure the accuracy of the compositions. In

most cases, the weight loss was found to be < 0.3 %. Occasionally, bulk compositions of

some of these alloys were confirmed using electron microprobe analysis (EMPA) or wet

chemical analysis (Table 3.1) and, as can be seen, the compositions the alloys are reliable.

Figure 3.1 is a graphic representation of the matrix of alloys, selected for this study.

3.2 Thermal and Thermo-Mechanical Processing

For the ALCHEMI and microhardness studies, each ingot was given a homogenization

treatment followed by step cooling, chosen to minimize thermal defects, which consisted of

annealing at 1573 K/5h + 1373 K/24h + 1173 K/48h + 1073 K/168h + furnace cool under

vacuum (Heat A). In some cases, this heat treatment was followed by a 1073 K/72h/WQ

(Heat B) or 1573 K/lh/WQ (Heat C) under vacuum. Other experiments were conducted

with heat treatments that differed in length and cooling rate from the three mentioned above

and are denoted as such in the results.

50% %Ni -

Binary or Ni-deficient
A Ni/AI = 1
o Al-deficient

% Fe

10% L- 10%
Ni + Fe = 50 Ni/AI =1 AI + Fe =50

Figure 3.1. Schematic representation of the matrix of alloys selected for this study.

% Fe

Selected alloys were examined using three-point bend testing. Specifically, 25mm

diameter ingots containing 0.125, 0.25, 0.5 and 1.0 % iron were encased in 75mm 0. D.

billets of 304 stainless steel and extruded at 1173 K to a cross-sectional area reduction of

9:1. The extrusions were then cut out from the canning material by electro-discharge

machining (EDM) and heat treated at 800C (1073 K) for 1 h. Specimens were cut to final

dimensions and electropolished. Finally, the bend bars were given a vacuum anneal at

900C (1173 K) for 1 h and cooled to room temperature by pulling the samples directly out

of the furnace hot zone without breaking the vacuum (i.e., vacuum cooling).

Table 3.1. Compositional analysis of selected alloys.

Element (at. %) Ni-50AI Ni-50A1-2Fe Ni-50Al-lOFe Ni-45A1-5Fe
Ni* 50.05 48.22 39.95 50.11
Al* 49.93 49.73 49.95 44.71
FetA 0 2.O1^ 10.08t 5.04t
Ct 0.0107 0.0216 0.0099 0.0114
N 0.0009 0.0086 0.0018 0.1138
O 0.0046 0.007 0.0083 0.0105
Analysis by wet chemistry/titration techniques (relative accuracy 1%)
t Analysis by wet chemistry/titration technique (relative accuracy 5%)
A Analysis performed on Shimadzu Ultraviolet/Visible Spectrophotometer, Model UV-160
(relative accuracy 10%)
t Analysis performed on LECO Corp. carbon/sulfur determinator, Model 244
(relative accuracy 10%)
Analysis performed on LECO Corp. nitrogen/carbon determinator, Model 436
(relative accuracy 10%)

3.3 Lattice Parameter Determination and Density Measurements

Lattice parameter measurements were determined by X-ray diffraction (XRD)

experiments on samples ground into -325 mesh (<45 gim) powder after homogenizing at

1573 K for 5h. The powder was then encapsulated in quartz with a titanium getter and

annealed at 1073 K for 45 h and cooled in air while the ampoule was still intact. The XRD

scans were run on a Philips 3720 diffractometer with generator settings of 40 kV and 20

mA. The scan rate was 0.8 deg/min and 0.1 deg/step over a range of 68 < 20 < 130

degrees. Standardized silicon powder (ao = 5.4301) was intermixed with the alloy

powders as an internal calibration. The function, sin20/cos0, for each peak was plotted

versus lattice parameter and an ao was extrapolated for a 20 value of 180 [see Cullity, 96].

Theoretical density values were calculated from the measured compositions and the lattice

parameters, assuming two atoms per unit cell.

Archimedes density measurements were conducted on samples that were crushed to a

particle size between 10 and 40 mesh (e.g., 2000 425 gim). The powders were taken

from the same ingots as those used for the lattice parameter studies. A 3 5 gram powder

sample was contained in a quartz vial and weighed in air and in a heavy liquid (n-

butylcellusolve). To measure the weight of the sample in liquid, the vial was first filled

with liquid, stirred to evacuated trapped air and the carefully submerged as to not loose any

powder. The sample was weighed before and after liquid submersion to check the validity

of the experiment. The temperature was monitored and the density of liquid was
determined from the known relationship between the liquid density (Pl) and temperature

[96]. The sample density (Ps) was calculated from the relation between the weight of the

powder in air (ma) vs. its weight in liquid (ml):

Ps =(ma/ma-mI) (PO)
These values were compared with calculated densities from the measured lattice parameters

and vacancy concentrations were then estimated.

Additionally, two XRD scans were made to measure the superlattice to fundamental

peak intensity ratios of the Ni-50A1 and Ni-50AI-10Fe samples. A count time of 8 s with a

step size of 0.2 deg (20) was performed over an angular range of 10-100(20). The

superlattice to fundamental peak ratios were calculated for each alloy and, based on the

different scattering factors for Al vs. Fe, the site occupancy of iron was inferred from this


3.4 Atom Probe Field Ion Microscopy

Specimens for atom probe field ion microscopy (APFIM) were prepared by cutting

rods that were 0.25 x 0.25 x 10 mm with a low speed diamond saw. One end of these

rods was mounted into a copper holder and electropolished using a two-stage polishing

procedure in a bench polishing system (Figure 3.1 a and b). First, a thin layer of 25%

perchloric acid 75% acetic acid solution, suspended on the surface of an inert liquid

(GaldenTM), was used to rapidly thin the middle portion of the rod without sharpening the

end by positioning the sample so only the middle was in contact with the electrolyte (Figure

3.2a). The voltage to the sample was maintained between 10 and 20 V. Once an hourglass

shape was formed, a second polishing stage was implemented. The sample was immersed

in a solution of N-butyl cellosolve and 2% perchloric acid and carefully electropolished at a

voltage of 6 12 V until the free end of the rod dropped off. This end is then retrieved

from the solution and mounted in a copper holder and, after examination repolished using a

loop shaped electrode in which a drop of the second stage polishing solution was

suspended. The sample was resharpened by inserting it through the loop and drawing the

tip into the solution as the voltage was turned off (Figure 3.2b). All stages of polishing

were observed through an optical microscope (30X) and the voltage was controlled to

maintain optimum polishing conditions.

The specimens of interest were examined in an energy-compensated atom probe (Figure

3.3 a and b) at Oak Ridge National Laboratory. The specimens were imaged after cooling

to temperatures between 30-60 K using neon as the imaging gas at a pressure of 2 x 10-5

torr. Atom probe analysis was performed at a pressure < 1 x 10-9 torr between 40 and

50 K by evaporating atom layers with a voltage pulse that was 20% of the total voltage to

minimize artifacts. Each specimen was analyzed to failure; in some cases, this took several

days and yielded data sets of several thousand atoms. For each alloy, 2 4 specimens were

analyzed depending on the length of each run. The applied voltage was periodically







Figure 3.2. Schematic illustrations of APFIM sample preparation in stage 1 and 2 (a)
and repolishing (b) [85].

--I -- i -- '| COMPENSATING
CHAMBER f-r "1 --- \



0 AI
A0 Ni



/oo ooooooo oo\


Illustrations of APFIM technique for imaging (a) and atom probe analysis
(b) [85].

Figure 3.3.

adjusted to achieve a controlled evaporation rate. The probe aperture was positioned

directly over the [001] pole. Site occupation information was determined by measuring the

compositions of the { 100} type planes which alternate between pure nickel and pure

aluminum in stoichiometric B2-ordered NiAl [87]. Iron was presumed to occupy a certain

site based upon the type of atoms evaporated before and after each iron atom that was

collected. Site occupancies (PFe'Ni') were determined for each alloy as the percentage of the

total iron atoms collected that occupied the nickel sites.

PFe'Ni' = [# Fe'Ni'/ (# Fetotal)]
The standard deviation (s) of each alloy was estimated as V/(PFe'Ni' # Fetotal).


Specimens were prepared by cutting and grinding the alloys of interest into 3 mm disks

that were =150 250 gtm thick. These disks were electrolytically thinned in a twin-jet

polishing system using a solution consisting of 6% perchloric acid, 34% N-butyl alcohol

and 60% methanol with electrical currents from 20 to 40 mA at temperatures from 233 to

253 K. ALCHEMI experiments were performed using a Philips CM 12 transmission

electron microscope, operated at an accelerating voltage of 120kV, equipped with an EDAX

9900 ultra-thin window energy dispersive X-ray spectrometer and a cryogenic specimen

holder at a temperature of 143 K. The specimen tilt geometry was fixed in relation to the

X-ray spectrometer in order to maintain a constant Al Kca absorption. The specimen was

masked with a gold washer (600 gtm aperture) to minimize secondary excitation of the

constituent peaks. A strain free area of desired thickness was selected based on the

intensity and sharpness of the Kikuchi pattern. The specimen was oriented so that only one

systematic row of reflections was strongly excited (g = 110 or 200). At least 10 spectra per

analysis were obtained from a single area of the specimen. The beam intensity was

monitored using the screen current meter and adjusted regularly to maintain a constant

level. The sample was tilted to minimize all non-systematic reflections, so that channeling

from only the planes of interest occurred. The intensities (Ii) of the Ni Ka, Al Ka and Fe

Ka peaks were determined by integrating over the peak area and subtracting the


The normalized intensities (ri = Ii/Li0) for Ni, Fe and Al ranging from strongly

channeling toward non-channeling conditions were plotted in relation to one another for

each diffracting condition. Figure 3.4a, b and c are representations of the normalized

intensities from ALCHEMI experiments for { 110} and {200} type planes, respectively.

The slopes (drNi/drFe and drAl/drFe), in Figure 3.4a, are the ionization delocalization

coefficients LNiFe and LAIFe. The difference between these two lines is a representation of

LNiAI which is defined as LNiFe/LAIFe. Seven data sets were collected from different

positions along the { 110} orientation. Average values for LNiAI and LNiFe were 1.379

0.009 and 0.996 0.007, respectively. These values were used to correct for

delocalization during ALCHEMI experiments using {200} systematic reflections.

Figure 3.4 b and c are representations of the normalized intensities (rNi, rAl, rFe)

plotted in three-dimensional space acquired from channeling on the {200} systematic row

of reflections. Instead of a line, as in the case of Figure 3.4a, the data is represented by a

plane. The planar coefficients aXFe'Ni' and cFeAI are obtained from the planar fit of the data

and the correlation coefficients (XNiFe and XFeAI ) can be obtained by correcting for

ionization delocalization.

If Ni and Al are perfect markers for their respective sites (i.e., PNiNi = PAIAI = 1),

XFe'Ni' directly corresponds to the fractional site distribution of iron on the nickel site

(PFe'Ni'). If Ni and Al are not perfect markers, the site distributions and constitutional
defects for each site can be determined, given the composition of the alloy (fNi, fFe, fAl)

and the character of the constitutional defects present. Since NiAl is a triple defect

compound [15], it is assumed that the primary defects present are VNi for Ni 5 50% and

NiAl for Ni 2 50%. The effect of anti-site concentration on site occupancy is discussed in

Chapter 2.


1.4 ]


I --- I -- ,^ -- I --- \- I --- I AI ,rpe
0.8 1.2 1.4 rAI re

] rAi 0 rFe
i Eli


Figure 3.4. ALCHEMI relations for determining the delocalization constant (a), and the
relationship between (rNi, rAj, rFe) viewing parallel (b) and orthoganonal to
the planar fit (c).


0 1.2




Figure 3.4. (continued)

1.2 1.4


0.4 0.6 0.8 1.2



3.6 Mechanical Testing

3.6.1 Bend Testing

Extruded rods (section 3.2) were then EDM'ed into bend test specimens in rough

dimensions of 2.25 x 2.25 x 12 mm, ground to 600 grit SiC paper and electropolished in a

13% H2SO4-77% methanol solution for 2 minutes at 7 volts. The final dimensions of each

bend bar varied slightly but typical values were 1.8 x 1.9 x 12 mm. Finally, these

specimens were heat treated to relieve stresses due to grinding (section 3.2) and a resulting

microstructure that was single phase with a 30 gJm grain size (linear intercept method) was


The alloys were tested in a three-point bending configuration with a bottom span of 10

mm on an Instron Model 4501, screw driven mechanical testing machine. The cross-head

velocity was constant at lim/nsec and the testing temperature was 295 K with 25% relative

humidity. Elastic equations were used to evaluate stress and strain relationships for each

alloy. The maximum stress and maximum outer-fiber strain were calculated using the

following relations:

a=l.5P1 / (wt2)

e=6td / 12

where t is the sample thickness, w is the sample width, d is the cross-head displacement, 1

is the span and P is the load. The 0.2% yield stress was determined by acquiring the load

(P) at the displacement (d) for a strain (-) of 0.002 from the load-displacement curve.

Likewise, the fracture stress and strain to failure were determined using the maximum load

at failure (Pmax) and the maximum displacement at failure (dmax) from the load-

displacement curve. The displacement values used were corrected for machine and

specimen compliance.

3.6.2 Microhardness

Microhardness specimens were cut from ingots for the three different heat treatment

schedules. These samples were set in cold-mounting epoxy, wet-ground using SiC paper

down to 600 grit size and polished with 0.3 gJm alumina. The heat treatment schedules

yielded large grain, single-phase microstructures. Each sample was tested at least 12 times

on a Buehler microhardness tester using a load of 300g for a duration of 15 s. At least

three different grains were examined in each specimen and an effort was made to avoid the

influence of grain boundaries by maintaining at least 1.5 times the indentation diagonal

distance from the boundary. Each indentation was examined for cracks or non-uniformities

and the results from irregular shaped or "bad" indentations were ignored. An average

microhardness number, H, was calculated with a standard deviation, s, as presented in

chapter 4.

3.7 Characterization

Fracture surfaces were examined using scanning electron microscopy. Fracture plane

normals were determined for single crystal fracture surfaces using photogrammetric

techniques [106]. By measuring a distance between features on a fracture plane and

observing the change in this distance over a known tilt radius, the angle of the fracture

plane with respect to the tensile axis was obtained.

Samples selected for analytical electron microscopy (AEM) were prepared using

techniques described for the ALCHEMI analysis. Deformation substructures were

examined and characteristic burgers vectors (b) were determined using standard diffraction

contrast g- b analysis. Line direction and character were determined using stereographic



The relationship between strength and iron content is known to differ with deviations

from the stoichiometric composition. To facilitate an understanding of the defects present

as a result of solute content, the site occupancy of iron and the nature of constitutional

defects in NiAl must be understood. The question of the site occupancy of iron in a series

of NiAl + Fe alloys is addressed initially. This is then related to the mechanical behavior of

these alloys in terms of site occupation and substitution scheme. This study will then

address the deformation mechanisms, ductility, solution strengthening and fracture

behavior of NiAlFe alloys.

4.1 Site Occupancy

The site occupancy of NiAIFe alloys was studied using three independent techniques:

(1) X-ray diffraction, (2) APFIM and (3) ALCHEMI. The alloy matrix selected for this

study is presented in Table 4.1 along with a summary of which of the three techniques was

used to determine the site occupancy behavior and the expected site preference of iron

based on previous studies.

4.1.1 X-ray Diffraction

X-ray diffraction measurements of selected alloys was conducted to obtain information

that would allow predictions concerning both site occupancy and lattice parameter of the

NiAl + Fe alloys. The lattice parameter (ao) was determined by XRD of annealed powders

intermixed with silicon "standard" powders. The resulting lattice parameters were then

Table 4.1: Matrix of Ni-Al-Fe alloys investigated along with the expected site preference of
iron based on previous work [11, 80, 98].

Substitution Scheme Composition Site Occupancy Expected Site
(at. %) Technique* Preference
Ni<50Al>5OFex Ni-50.7Al-2.3Fe 2, 3 Ni
Ni-50A1-0.25Fe 3 Ni
Ni50-xAl50Fex Ni-50AI-2.0Fe 2, 3 Ni
Ni-50A1-5.0Fe 1,2, 3 Ni
Ni-50AI-10OFe 1,2, 3 Ni
Ni-49.875A1-0.25Fe 3 Ni or Al
Ni50-xAl50-xFe2x Ni-49A1-2.0Fe 3 Ni or Al
Ni-47.5A1-5.0Fe 3 Ni or Al
Ni-45A1-lOFe 3 Ni or Al
Ni-49.75A1-0.25Fe 3 Ni or Al
Ni50Al50-xFex Ni-48A1-2.0Fe 3 Ni or Al
Ni-45A1-5.0Fe 3 Ni or Al
Ni-40Al-10Fe 1,2, 3 Ni or Al
Ni>50Al<5OFex Ni-48.5A1 -0.3Fe 2, 3 Ni or Al
1. XRD (compared with Archimedes density measurements)

combined with the composition of the alloy to yield a theoretical density (based on an

assumption of 2 atoms/unit cell); by comparing it with an experimentally measured density,

an estimate of vacancy concentration was obtained. A summary of the vacancy

concentrations estimated using this technique is given in Table 4.2. The primary source of

error in this calculation comes from the density measurements, which is estimated to be

0.3%. It is not surprising that the only aluminum-rich composition examined has the

highest vacancy concentration (VNi=1.43%). Also, both binary NiAl and Ni-40A1-10Fe

yield negative numbers suggesting that essentially no vacancies are present in these alloys.

However, a trend appears for nickel-deficient alloys in which the vacancy content is

increased as a result of iron additions. Whether these are residual thermal vacancies that

have been stabilized by iron or constitutional vacancies generated by a deviation from the

expected site occupancy is not known. It is unclear how the vacancy concentration relates

to iron content since Ni-50AI-5Fe has a higher apparent vacancy concentration than Ni-

50A-10OFe. It should be noted, however, that these values could be within the expected

scatter since scatter in both the composition and density measurements is 0.3 %.

Table 4.2. Lattice parameter and Archimedes density measurements
for iron-containing NiAl alloys (see Appendix A).

Alloy Lattice X-ray Archimedes Calculated
(at. %) Parameter Density* Density Vacancies+
_________ ao (A) (g/cc) (g/cc) (%)
Ni-50AI 2.8871 5.9117 5.9191 -0.13
Ni-50.7AI-2.2Fe 2.8874 5.8702 5.7861 1.43
Ni-50A1-5Fe 2.8878 5.8878 5.8252 1.06
Ni-50A1-lOFe 2.8881 5.8663 5.8484 0.31
Ni-40A1-10Fe 2.8800 6.3568 6.3613 -0.07
* Calculated from XRD analysis, assuming 2 atoms/unit cell
+ Percent of the total number of sites that are unoccupied

Another method for site occupancy determination with X-ray diffraction involves

estimating the character of atoms on a given site from the measured ratios of the integrated

peak intensities of superlattice and fundamental peaks, Is/If, and comparing those to

theoretically calculated values Iscalc/Ifcalc based on the known scattering factor of each

element [100]. In the NiAl + Fe system, however, such measurements will not distinguish

between Ni and Fe due to their similar scattering factors. On the other hand, it is possible

to use this ratio to determine whether Al atoms occupy nickel sites or Ni or Fe occupy

aluminum sites. In the case of Ni-50AI-lOFe, if iron occupies the nickel sites exclusively,

these calculations give a ratio value (Is/If IfCalc/Iscalc) of 0.99 compared to 0.39 if all the

iron displaces aluminum onto the nickel sites. It is estimated that 1% VNi'S reduces this

ratio from 0.99 to 0.89 while 2% reduces it to 0.83. Figure 4.1 a and b show the spectra

for Ni-50Al and Ni-50Al-lOFe, respectively. When the intensity ratio was measured for

Ni-50Al and Ni-50Al-lOFe, values of 0.95 and 0.88 were observed, respectively.

Although the value for Ni-50A1 is slightly lower than that calculated (0.95 as opposed to

1), it is considered to be well within the expected scatter. The value for Ni-50A1- lOFe

deviates even more from unity but, as shown above, this could be explained by the

presence of a small amount of VNi' s.

The X-ray data obtained from the Fe-doped alloys suggest that iron occupies nickel

sites when substituted for nickel. Small amounts of vacancies are produced by either some

of the iron or nickel occupying aluminum sites (< 1% of total atoms) or an increased

thermal vacancy concentration stabilized by the addition of iron. When iron is substituted

for aluminum, either it or some of the nickel will occupy aluminum sites, i.e., no vacancies

are created as a result of alloying in this substitution scheme.

4.1.2 APFIM

A series of NiAl alloys containing iron were examined using atom probe field ion

microscopy. In Figure 4.2, field ion micrographs of Ni-48.2Al-O.3Fe and Ni-50.7A1-

2.2Fe show the type of images observed. Dark regions on these images are caused by

preferential evaporation of atoms in certain crystallographic directions; dark and light layers

of atoms near the [100] pole correspond to Al and Ni layers (arrows in Figure 4.2),

respectively. A typical sequence of evaporation is illustrated in Figure 4.3, where the

characters "*, a, and F" denote nickel, aluminum and iron atoms, respectively. As can be

seen in this figure, the site occupancies of iron are determined by their association with a

large group of nickel or aluminum atoms. Each iron atom is assigned to one of five

categories based on its position in the evaporation sequence: atoms occupying aluminum

sites (Al), nickel sites (Ni), the end of each respective plane (eAl, eNi) or unknown (unk).

Since the evaporation potential of iron is higher than that of aluminum, it is expected that

iron would tend to evaporate last on aluminum planes and, hence, it was assumed that iron



0 !



X-ray diffraction spectra of NiAl (a) and Ni50A110Fe (b)

Figure 4.1.



t 10000-



10 20 30 40 50 60


70 80 90 100

Figure 4.1. (continued)


sitting on eAl sites corresponded to iron on Al planes. The evaporation potentials of iron

and nickel are very similar so the same assumption cannot be made in the case of eNi.

Thus, for these experiments, iron atoms found to occupy eNi sites were treated as

unknown (unk) and not factored into the probability calculations. In this case, it is clear

that the majority of iron atoms evaporated while occupying the aluminum planes. In Figure

4.4, a compositional profile is shown for Ni-50.7A1-2.2Fe where it is apparent that iron

appears to favor the Al-sites exclusively.

The measured number of iron atoms occupying nickel sites in each alloy is presented in

Table 4.3 along with the total number of iron atoms collected. The standard deviation

represents only the statistical error in the experiments and no systematic contributions to the

error were considered. In all five alloys, both aluminum-deficient and nickel-deficient

cases, iron was observed to prefer aluminum sites. Specifically, in every case but one,

iron was shown to exhibit a 90% or higher probability of aluminum site occupancy. In the

case of the 5% Fe alloy, only a 72% probability of aluminum site occupation was

measured. It is not clear why this alloy exhibits a lower site preference for aluminum than

the other nickel-deficient alloys.

Estimates of constitutional defects, as calculated from the experimental data, suggest

unrealistic defect concentrations that were inconsistent with the XRD results. Therefore, it

was assumed that some sort of systematic errors associated with this experimental

technique produced these discrepancies. However, it should be emphasized that not all of

the alloys characterized by APFIM yielded anomalous results, i.e., the two aluminum-

deficient alloys showed a low probability for iron on nickel sites, in agreement with

previous predictions.

(a) .*. .


Figure 4.2. Field ion image of NiAl containing 0.3 % Fe (a) and 2.2 % Fe (b). On
(100) planes bright layers are nickel and dark are aluminum.

................. aaaaaaaF
aaaFaa ...................
* aaa*a*aaaaaaaaaaaa .....
.................. aaaaFFa
. Fa ......................
. aaaaFFaaaaaaa......
...................... FaF
aaFFFaaa-aa ..............
.......FaaFaaaFaaaFaaaa. F
Faa ......................

_____Fe site occupancy
Al Ni eAl eNi Unk

1 1
4 1

Figure 4.3. Typical evaporation sequence of an NiAIFe alloy (Ni-40AI-10OFe)
with "-, a, and F" referring to Ni, Al and Fe, respectively.

Visual evidence of transition metal anti-site atoms present in an aluminum-deficient

alloy are shown on Figure 4.5. Arrows denote the high contrast from brightly imaging

atoms which have a high evaporation potential (e.g., transition metals) in the aluminum

plane. This suggests that transition metal anti-site atoms (i.e., nickel or iron on aluminum

sites) are present. Thus, if iron occupied the aluminum plane in other alloys, it would be

visible as well in the atom probe. Yet, the alloys that contain < 50% nickel do not show

this contrast in the aluminum plane. Therefore, it appears that the transition metals do not

occupy aluminum sites in the nickel-deficient alloys largely because of the energy required

to form VNi'S becomes a dominant factor.

In order to understand the origin of these systematic errors, it is necessary to

understand the methods used to acquire the data. In Figure 4.3, it can be seen that the

evaporation sequence is not always uniform. Specifically, nickel atoms will occasionally

be observed in the middle of aluminum planes and vice-versa. This problem arises because

the rate of evaporation for aluminum is significantly greater than that for nickel (or iron);

hence, nickel (or iron) atoms may be retained on the surface while the aluminum atoms

evaporate from the plane underneath. This can be understood when considering the

relationship between applied potential and evaporation rate in Figure 4.6a. As this

schematic shows, the evaporation rate of a layer is proportional to the amount by which the

voltage pulse exceeds the constituents' evaporation potentials. An evaporation rate for

nickel of 10-2 atom layers/s may result in an evaporation rate of 106 atom layers/s for

aluminum. Since the evaporation field potential for aluminum (FAI = 19 V/nm) is

approximately half that for nickel or iron (FNi = FFe =35 V/nm), the aluminum atoms may

come off irregularly and fail to be counted. This causes the overall composition of the

collected atoms to be aluminum-deficient compared with the actual alloy composition and

could cause the irregular evaporation sequence illustrated in Figure 4.6b. A photographic

sequence is presented in Figure 4.7, where arrows denote aluminum atoms in the second

layer that evaporate before the last transition metal atoms in the layer above. Since

uncontrolled evaporation of the aluminum atoms often took place during the experiments, it

is difficult to predict the actual site occupancy of iron from this technique.

Table 4.3. Site occupation behavior for iron in NiAl alloys as determined by APFIM.

Alloy PFe'Ni' Expected Defects Number of
(atomic %) (% s) VNi, NiAl or A1Ni Fe atoms
(%) collected
Ni-50.7A1-2.2Fe 7 2.9 5.46 (VNi) or 74
2.75 (A1Ni)
Ni-50A1-5Fe 28 4.6 7.2 (VNi) or 95
3.6 (A1Ni)
Ni-50Al-10Fe 3 1.0 19.4 (VNi) or 285
9.7 (A1Ni)
Ni-48.5A1-0.3Fe 9 5.9 2.45 (NiAl) 23

Ni-40AI-10Fe 6.7 6.3 1.34 (NiAl) 60

It should be noted that the evaporation field potentials quoted above were calculated

from the image hump model [85] and are values for these elements themselves. This

model does not take into account bonding arrangements or character. Hence, the actual

FNi, FFe or FAI values may differ drastically from the predicted values. Since an atom



20 -

20 30
(001) PLANE

Figure 4.4. Atom probe composition profile for Ni-50.7A1-2.3Fe in which
iron is found on aluminum sites.


Figure 4.5. A Field Ion Micrograph of an aluminum {200} plane in a
Ni-40A-10OFe alloy, showing the transition metal atoms on the aluminum

must break bonds to become ionized and evaporated, it is relatively straightforward to

expect the nearest and next nearest neighbor bond strengths to play a role in the expected

evaporation potential. Because Al acts as an electropositive element in NiAI, the A1-A1

interactions have been predicted to actually be repulsive in nature [37]. Therefore, it is

reasonable to assume that the stability of an aluminum plane of atoms in NiAl is much

lower in the presence of an electric field than that of a nickel plane such that the probability

of aluminum evaporation prior to the completion of the previous nickel layer is good. In a

similar manner, the iron and/or nickel atoms can be retained from the previous nickel layer

and evaporated toward the end of the aluminum plane.

Based on this complication, the conditions in previous studies merit discussion when

considering the viability of this technique. In most cases, the difference in evaporation

potential between alternating layers is not as drastic. In the case of Ni3Al + Hf, Co and Fe

[86], alternating layers vary in composition between all nickel and 50/50 nickel/aluminum

such that the average evaporation rates for a given layer would not vary as dramatically.

For higher Al phases (e.g., XAl), most of the studies to date have been on Al-deficient

alloys in which the ternary alloying addition substitutes primarily for aluminum. For

example, the FeAl + Cr system studied by Kosla et al. [83] contained at least 60% Fe and

5% Cr, such that =30% of the aluminum sites were occupied by transition metals which

would tend to stabilize the aluminum layer during evaporation. In the case of NiAl + Be,

the alloy characterized had very low concentrations of Be and all of this was found to

occupy aluminum sites. Hence, the previous studies in this field have not dealt with such a

large variation in the evaporation potentials and rates, or with a range of compositions.

Solutions for eliminating this systematic error can be surmised from Figure 4.6. Either

a drop in temperature or an increase in the voltage pulse would cause the atomic layers to

evaporate at a more uniform rate. In this study, unsuccessful attempts to stabilize the

evaporation rate by varying temperature and pressure (also known to effect the evaporation

Field Evaporation Rate(atom layers/s)

S- Fe


10^ 2 ___ Al

T1 T2



TI: Preferential Evaporation of Al
T2: Ideal Conditions
T3: Uncontolled Evaporation of Al


Fe Ni

Figure 4.6. Schematic of the influence of field strength on evaporation rate (a)
and field strength and temperature on the evaporation
behavior in an NiAlFe alloy (b).


\' "C




Field ion micrograph sequence (a-h) illustrating a tendency for aluminum to
evaporate from "underneath" nickel or iron atoms in Ni-50A1-5Fe.

Figure 4.7.

potential) were made. Because these attempts were unsuccessful, it is believed that only an

increased voltage pulse, above the 20% possible with the current APFIM would allow for

thorough examination of this system by atom probe. However, investigations involving

the implementation of a more powerful pulse generator may be undertaken in the future.


ALCHEMI experiments were conducted to determine the site occupancies of iron in NiAl.

Figure 4.8 is an example of spectrum acquired from this technique and shows the effect of

orientation on the characteristic peak intensities of nickel, aluminum and iron. The

normalized intensities of these peaks, rAI- 1, rNi- 1, rFe- 1, when corrected for delocalization,

LAI, LNi, LFe, are plotted as a function of orientation (see figure 4.9a and c) in order to

illustrate this relationship when iron occupies different sites1. Corresponding

microdiffraction patterns are also presented for the orientations used in these two

experiments (Figure 4.9 b and d). These plots clearly show that rFe can be related directly

as some combination of rNi and rAl. Specifically, in Figure 4.9a, rFe corresponds directly

with rNi which suggest that iron and nickel occupy the same sublattice. Alternatively, in

Figure 4.9c, rFe corresponds more closely with rAI than rNi, suggesting that the majority of

iron occupies aluminum sites. When a plane is fit to the data (see section 3.5), the site

occupancy for Fe on Ni sites, PFe'Ni', is obtained from the planar coefficient, aFeNi (when

corrected for delocalization). Initially, samples from the same alloys and heat treatments as

the X-ray and the APFIM studies were used to assure a direct comparison between the

techniques. These preliminary results are presented in Table 4.4 where it can be seen that

PFe'Ni' for the nickel-deficient alloys contradict the data obtained in the APFIM analysis.
Furthermore, the vacancy estimates from the Ni-50.7A1-2.2Fe alloy correspond closely

with the VNi estimates obtained from the XRD/density measurements. The alloys in which

1 Graphic formulation by I. M. Anderson in unpublished research, 1995.

0/0B = 2.8

r=(1, 1, 1



0 1 2 3 4 5 6 7 8 9 10
AlKa K /OB = 0.9

r= (1.01, 1.43, 1.43)




0 1 2 3 4 5 6 7 8 9 10

AIKa 0/0B = 1.25
r = (0.77, 0.89, 0.80)



0 1 2 3 4 5 6 7 8 9 10

O/OB = 0 (symmetry)
r = (1.55, 0.91, 1.02)



0 1 2 3 4 5 6 7 8 9 10

Figure 4.8. EDS X-ray sprectra for Ni-40AI-10OFe alloy at various tilt orientations (O/OB) with normalized intensities r = (rNi, rAl, rFe)
given for each condition.



L[Ni,X] (r[X] 1)




0.5 1 1.5 2I .5



Figure 4.9. Normalized intensity (rNi-1, rAl-1, rFe- 1), corrected for delocalization
([Lx][rx- 1]) vs. channeling condition for Ni50A15Fe (a) and Ni45A15Fe (c)
with corresponding microdiffraction patterns for selected tilting conditions
(b and d, respectively).

Figure 4.9. (continued)

L[Ni,X] (r[X] 1)


0.6 N 0



0.5 1.5g
-0.2 \

Figure 4.9. (continued)

Figure 4.9. (continued)

iron was substituted for both Ni and Al (Ni/Al = 1) or for Al only show that iron prefers

aluminum sites as long as it does not result in the formation of a large number of VNi' s.

Table 4.4. Site occupancy of iron in NiAl alloys determined from ALCHEMI

Alloy Composition Measured Site Occupancy Estimated Constitutional
(substitution scheme: PFe'Ni' Defects: NiAl or VNi
Fe for Ni or Al) (% s) (% s)
Ni-50Al-10Fe* (Ni) 88.0 1.8 2.4 0.4 (VNi)
Ni-50.7Al-2.2Fe+ (Ni) 94.1 3.2 1.7 0.1 (VNi)
Ni-49.75Al-0.5Fe+ (Ni) 54.5 3.2 0.0 0.1 (NiAI)
Ni-40Al-10Fe* (Al) 25.3 1.5 2.6 0.2 (NAI)
* 1573 K/5h + 1073 K/12h
+ 11373 K/lh + extrude 9:1 + 1173 K/lh + 1073 K/lh

Following the initial set of experiments, a larger matrix of alloys, water quenched from

1073 K to prevent any possibility for precipitation in the iron-rich alloys, was examined.

The planar coefficients from this matrix of alloys are tabulated in Appendix B; the results

are presented in Table 4.5 and represented graphically in Figure 4.10. The standard

deviations represent the measure of deviation from the planar fit and are inversely

proportional to the square root of the iron intensity. As expected, the site occupancy

behavior of iron in NiAl differed with stoichiometry and, as seen in Figure 4.10, iron

exhibited a preference for whichever site minimizes the constitutional defects, i.e., if

substituted for nickel, iron favors Ni-sites and vice-versa. In the nickel-deficient alloys,

iron was determined to occupy nickel sites almost exclusively. However, even a small

number of iron atoms occupying aluminum sites will give rise to vacancies on the nickel

sublattice in these compositions.

In Table 4.5, the concentration of iron atoms which occupy the nickel sublattice is

included along with the concentrations of constitutional defects predicted from the site

occupancy. Using the XRD results (section 4.1.1) for the nickel-deficient alloys, the

constitutional defects generated as a result of iron's site occupancy were assumed to be

VNi'S. For the other two substitution schemes, NiAl's were assumed to form. These

values can be used as indicators for properties. For example, because vacancies strengthen

NiAl at a rate of more than three times that of nickel anti-sites, vacancy-rich compositions

should exhibit higher strengths than expected if PFe'Ni' was 100%.

As previously mentioned, Fu et al. [80] used first principle calculations to model the

predicted site occupancy of iron in NiAl as a function of solute level, stoichiometry and

temperature. When iron was substituted for aluminum, they predicted it would begin to

prefer the nickel sublattice with decreasing concentration. The ALCHEMI data also

exhibits a similar preference; i.e., as the solute level decreases, PFe'Ni' increases from

approximately 30% to 55%. Fu et al. also predicted that iron substituted for nickel always

resulted in PFe'Ni' of 100% regardless of the temperature or solute level whereas they

predicted rapid vacancy production for aluminum concentrations above 50%. Although the

data in this study does not quantitatively agree with this model, the general trends are


4.1.4 Summary of Site Occupancy Studies

Of the various studies dealing with site occupancy in the literature, the present research

appears to have been the most comprehensive in terms of the large matrix of compositions

in which site occupancy was determined as a function of both solute level and substitution

scheme. Previously, it has been difficult to examine how experimental results compare

with the theoretical models of site preferences in ordered compounds. With the exception

of the study by Kolsa et al. [83], most of these involved one or two alloy compositions and

differences in stoichiometry. Few studies have compared the results from different

techniques to insure the absence of systematic errors in the data. Prior to this study, the

method of lattice site occupancy determination by APFIM was not reported to exhibit such

large systematic errors. This clearly shows that a deeper understanding of evaporation

ALCHEMI results for NiAlFe alloys*







2 4 6 8

*water quenched from 1073 K



% Fe

Figure 4.10. Iron site occupancy (PFe'Ni') as a function of solute level. (Ni,Fe)Al
denotes iron substituted for nickel, (Ni, Fe)(Al, Fe) denotes iron substituted for both nickel
and aluminum and Ni(AI, Fe) denotes iron substituted for aluminum.

Table 4.5: Site occupation of iron in NiAl alloys determined from ALCHEMI.

Site Constitutional Constitutional Constitutional
Alloy Composition Occupancy Fe Defects on Vacancies Anti-sites
(substitution scheme: PFe'Ni' Ni (VNi) (NiA)
Fe for Ni, Al, or both) (% s) (FeNi) (% s) (% s)
1 1 (% s)
Ni-50A1-0.25Fe (Ni) 91.6 8.7 0.229 0.022 0.042
89.1 8.1 0.220 0.020 0.06
Ni-50A1-2.0Fe (Ni) 88.4 2.9 1.77 0.058 0.464
Ni-50A1-5.0Fe (Ni) 96.4 0.9 4.82 0.045 0.360
Ni-50Al-10OFe (Ni) 95.3 0.7 9.53 0.070 0.94
Ni-49.875A1-0.25Fe 62.6 6.0 0.157 0.015 0.032
(both) 63.1 3.4 0.158 0.009 0.033
Ni-49AI-2.0Fe (both) 58.4 2.7 1.17 0.054 0.168
Ni-47.5A1-5Fe (both) 57.7 1.0 2.89 0.050 0.385
Ni-45Al-10OFe (both) 56.7 1.3 5.67 0.130 0.670
Ni-49.75AI-0.25Fe 49.4 3.9 0.124 0.01 0.123
(Al) 61.1 2.1 0.153 0.005 0.153
Ni-48A1-2.0Fe (Al) 29.0 1.6 0.580 0.032 0.580
33.5 2.0 0.670 0.040 0.670
Ni-45A1-5.0Fe (Al) 27.2 1.8 1.36 0.09 1.36
Ni-40AI-10Fe (Al) 23.7 0.9 2.37 0.09 2.37

kinetics should be obtained for compounds which consist of elements with significantly

different evaporation potentials.

From this study, it is apparent that iron exhibits a preference for aluminum sites in the

NiAl structure. In nickel-deficient alloys where iron is substituted completely for nickel

(i.e., (NilXFex)Al), this preference produces a small amount of VNi'S. The energy of

formation of a triple defect has been calculated to be 183 kJ/mole [15]. Since, in an alloy

which has 50% aluminum, a vacancy must be produced whenever iron occupies an Al-site

(i.e., FeAI), a large amount of energy is associated with every VNi. Although iron prefers

to occupy aluminum sites, only a small percentage of it does. Nonetheless, this

experimental study shows that a small portion of iron occupies aluminum sites in spite of it

being substituted for nickel.

In the case of aluminum-deficient alloys, this preference is still observed. If Ni and Fe

had an equal affinity for aluminum sites, the portion of FeAl'S and NiAl's would be

proportional to the relative fractions of iron and nickel in the alloy (i.e., 1/200 > Fe/Ni 2

1/4). Instead, the opposite is true, since the FeA1/NiAl ratio ranges between 2/3 and 6.2/1;

this is consistent with the previous literature where iron was predicted to favor aluminum

sites based on quasi-chemical models [77, 79] and first principle calculations [80]. It has

also been predicted from experimental methods that some, if not all, of the iron would

occupy the aluminum sites when substituted for aluminum [63, 99]. From this study, the

significance of this site preference is determined to be significant but not so drastic that iron

would occupy aluminum sites completely.

4.2 Mechanical Properties

Various aspects of the mechanical behavior in NiAl + Fe alloys have been studied in

light of the site occupancies determined in section 4.1. Slip behavior, ductility, fracture

behavior and microhardness were examined as a function of iron content, heat treatment

and substitution scheme for selected compositions. This section attempts to correlate the

mechanical behavior with the lattice defects caused by iron additions.

4.2.1 Slip Behavior in NiAl and NiAl + Fe Single Crystals

A comparison of dislocation substructures was conducted on two NiAl single crystal

alloys. Stoichiometric NiAl and Ni-49.75A1-0.25Fe were selected because of the different

ductilities exhibited in the literature [5]. Previously tested single crystal tensile specimens,

oriented in the [110] direction, were obtained from the study conducted by Darolia et al.[5].

The specimens were tested to failure at a strain rate of 1 x 10-3 s-1 at room temperature.

The strain to failure (ef) of the NiAl and Ni-Al-Fe specimens were approximately 1% and

6%, respectively. TEM analysis of these specimens were examined to determine if there

was any difference in character of the alloy with 0.25 % iron. In both specimens, the

dislocations appear to be beginning to form cellular morphologies (Figure 4.11). Although

the iron containing crystal had undrgone 6% plastic strain, the dislocation density was still

very low. Standard g. b analyses were conducted and, of the dislocations examined, all

were determined to have <100> type burger's vectors (See Appendix C). Hence, there

was no discernible change in the character of slip as a consequence of microalloying with

iron. Unfortunately, the slip plane was not as apparent. Appendix C shows a typical

stereographic projection from a slip trace analysis showing the line direction and slip plane

of a dislocation in NiAl. Some, but not all dislocations were found to be on {001 } planes

which is consistent with the data reported by Field et al. [26] for binary NiAl in { 110}

orientations. Field et al. also noted slip lines that were irregular and indistinct; this is not

surprising and suggests that dislocations in NiAl may cross-slip to many different slip

planes as in bcc metals. This { hk0 } slip has been reported in other studies as well but is

not seen as an effective mechanism of ductility enhancement. It is thought that NiAl single

crystals need multiple slip systems to overcome local pile-ups at defects. Only the scew

component of dislocations is able to cross-slip, leaving the edge component pinned at the


4.2.2 Ductility and Flow in NiAl + Fe Polycrystals

The effects of iron on the mechanical properties of polycrystalline NiAl have been

studied. The microstructures produced for each alloy are presented in Figure 4.12. Three

point bending was used to survey the properties of NiAl containing 0.125, 0.25, 0.5 and

1.0 percent iron, substituted for nickel and aluminum equally (Ni/Al = 1). The results of

the bend tests are summarized in Table 4.6. The average values of the 0.2 % yield stress,

maximum fracture stress and the strain to failure (maximum outer fiber strain at failure) are

Figure 4.11.

Dislocation structures for NiAl (a) and Ni-49.75A1-0.25Fe (b) single crystal
[110] tensile specimen.

given for each alloy along with appropriate standard deviations. The data are compared

with values for stoichiometric NiAl tested in three-point bending at a strain rate of 10-3 s-1

by Schneibel et al. [100]. It is noted that the thermal histories of the NiAl samples used by

Schneibel differed slightly from that used in the present study in that they heat treated their

alloy for 2 h at 1273 K and vacuum cooled to produce a single phase microstructure with a

grain size of approximately 35 gim.

The yield stress was calculated using a linear elasticity assumption which may result in

values that vary from the actual values measured in simple compression tests. Schneibel et

al. [100] used this method to estimate an average bending yield stress of 232 9 MPa for

stoichiometric NiAl while compression testing of the same extrusion resulted in a yield

stress of 241 21 MPa suggesting that the elastic estimate of yield stress lies within the

scatter of the compression data. Difficulties are also encountered when attempting to

compare the mechanical properties of NiAl alloys from the literature because of their

sensitivity to deviations from stoichiometry, impurities and processing conditions. In this

study, however, a direct comparison is more reasonable because the alloys were made from

the same starting materials, processed by the same methods and tested on the same testing


Table 4.6. Mechanical properties of NiAlFe alloys obtained from 3-point bend tests.

Alloy Yield Stress Fracture Stress Strain to Failure Number of
I yys (MPa) f-fs (MPa) Es (gm/gim) Samples Tested
NiAl* 2329 45626 0.01610.0028 12
NiAl-0.125Fe 2357.2 47041.0 0.01900.0042 6
NiAl-0.25Fe 2208.9 48143.5 0.02350.0053 6
NiAl-0.5Fe 23512.3 51929.4 0.02620.0033 6
NiAl-1.OFe 2398.2 46633.7 0.02020.0027 6
*J.H. Schneibel etal. [100]

Typical load-displacement curves for NiAl + Fe alloys are shown in Figure 4.13. The

curves appear similar to those in the literature for stoichiometric NiAl [100].

Unfortunately, the mechanical properties measured from bend tests are very sensitive to

surface defects, as is the case with tensile tests, so data acquired by this method can exhibit

scatter. In order to get an idea of the reliability of these bend tests, Weibull statistics [101]

were employed for the yield and fracture stresses of all the NiAl + Fe alloys (see Figure

4.14). Although the sampling size is relatively small for this method of error analysis,

Weibell moduli do indicate the reliability of the data. The Weibull modulus (m) is defined
as the slope of the ln(ln(I/P)) vs. ln(a) plot, where P is the probability of survival, and a is

the stress (yield or fracture). As shown in Figure 4.14a-d, the NiAl + Fe alloys exhibit

Weibull moduli that are fairly consistent with semi-ductile materials Because yield stress is

less sensitive to surface flaws, it should have a higher Weibull modulus than the fracture

stress. The Weibull moduli for all alloys are listed in Table 4.7 For fracture stresses, these

moduli are comparable to the values of many ceramic materials consistent with the low

fracture toughness (Kic) of NiAl. Both values are similar to the values obtained by

Schneibel et al. [100] for NiAl with a 600 grit surface finish. Values observed were 25.6

for my and 16.3 for mf. Thus, no significant increase in Weibull moduli are observed due

to the iron additions.

Table 4.7: Weibull moduli for NiAlFe alloys

Alloy Weibull Modulus for Weibull Modulus for
11 Yield Stress (my) Fracture Stress (mf)
NiAl-0.125Fe 29.7 10.4
NiAl-0.25Fe 22.9 19.0
NiAl-0.5Fe 16.3 113.5
NiAl-1.OFe 26.0 12.1

Figure 4.15 illustrates the mean value of the mechanical properties listed in Table 4.6 as

a function of iron content. As can be seen from the data, the yield behavior is reasonably

similar among all the Fe-containing alloys and appears comparable to that of NiAl. NiAl +

Figure 4.12.

Microstructures of the NiAl + Fe alloy extrusions: NiAl-0.125Fe (a),
NiAl-0.25Fe (b), NiAl-0.5Fe (c), NiAl-IFe (d).

Figure 4.12. (continued)


0 50 100 150


250 300


0 50 100 150 200 250 300 350
displacement, um

Figure 4.13

Typical load vs. displacement curve for NiAI+Fe alloys: NiAL-0.125Fe (a),
NiAl-0.25Fe (b), NiAl-0.5Fe (c), NiAl-1.OFe (d).















50 100 150 200 250
displacement, um

(c) 250 1 ---

200 -

z 150 -

S100 -

50 /

0 50 100 150 200 250 300
displacement, um


Figure 4.13. (continued)



1 -










-2 L
b 4.8


5 5.2 5.4 5.6

5.8 6

Figure 4.14.

Weibull plots of the yield and fracture stresses from 3-point bending of
NiAl + Fe alloys: NiAl-0.125Fe (a), NiAl-0.25Fe (b), NiA1-0.5Fe (c)
NiAl- 1.OFe (d) (P = probability of survival and m = modulus).

5.6 5.8 6 6.2








2 L
C 4.8


5 5.2 5.4 5.6

5.8 6


5 52

5.4 5.6

Figure 4.14. (continued)









0.25Fe exhibits a slightly lower yield stress than the other alloys but this difference is only

slightly larger than 1 standard deviation and is not considered significant. It should be

noted, however, that Darolia et al. [5] reported a drop in yield stress with iron additions

although a different substitution scheme was used in their alloys. There is also an apparent

increase in the fracture stress accompanied by a corresponding rise in the strain to failure as

iron is increased from 0.125 to 0.25 %. Interestingly, Noebe etal. [70] reported an

increase in yield stress from 180 to 203 MPa in polycrystals, as a result of 0.1% Fe added

to NiAl (substituted for Al); this was attributed to solid solution hardening. These results

complement this study but are not directly comparable because of the different substitution

scheme and possible differences in purity. They also do not appear to be comparable to

each other since solution strengthening and loss of ductility were observed in the

polycrystals while softening and increases in ductility were observed in the single crystals.

However, it does appear that the effect of dilute iron additions on polycrystalline NiAl is

small in comparison to its effect on NiAl single crystals [5].

4.2.3 Fracture Behavior of Dilute NiAl + Fe Alloys

Stoichiometric polycrystalline NiAl is usually observed to fail primarily by intergranular

fracture at room temperature. This has been reported to occur as a result of NiAl's failure

to meet the von Mises' criterion and its inherently weak grain boundaries [18, 20]. The

fracture behavior of the iron-modified NiAl bend specimens was studied to establish the

mode of fracture in these alloys and to determine if iron has any effect on fracture behavior.

In Figure 4.16a, b, c and d, SEM micrographs for NiAl + Fe alloys show that the fracture

surfaces are almost completely intergranular, suggesting that iron does not change the

mechanisms controlling fracture in polycrystalline NiAl. This in not surprising when

considering that all other studies which managed to suppress intergranular failure did so by

either suppressing the ductility or adding elements that segregated to grain boundaries

---- Yield Stress
-- o-- Fracture Stress
550 -- _,

450 -


CD 250 g

150^^^ I-

50 ----------
0 0.2 0.4 0.6
% Fe

Figure 4.15.

- Strain to Failure




0.06 "


j i 0
1 1.2

Yield stress, fracture stress and strain to failure as a function of iron
concentration in NiAl polycrystals.

(e.g., B or Zr). Iron is not expected to segregate to grain boundaries and, hence, should

only affect the fracture mode indirectly, i.e., by changing the deformation substructure and,

in effect, lower the stress concentrations at grain boundaries. Interestingly, the room

temperature fracture mode of NiAl + Fe polycrystals doesn't change even up to iron levels

of 10% although a loss of ductility prior to fracture is observed after only 2% 1.

In order to examine the effect of iron on fracture in single crystals, tensile specimens

from the study by Darolia et al. [5] were examined. The testing procedures are outlined

earlier in this section. The favored cleavage plane has been calculated as { 110} [37] but, in

tension and certain bend test orientations, the fracture plane has been observed to exhibit

cleavage on { 115} or { 117} type planes [38, 102]. Because this change in fracture mode

is poorly understood, it was hoped that any differences in the behavior of the iron-

containing crystals might elucidate the mechanisms involved. Figure 4.17a and b shows

the apparent points of initiation for fracture arising from surface defects. Whereas the

initiating defect is clearly a scratch, it is difficult to determine the nature of the defect in the

iron-containing crystal. In Figure 4.18a and b, overall views of the fracture surfaces are

shown along with stereographic projections, where the cleavage plane normals are shown.

All of the cleavage planes appear to have normals close to <115>. However, the fracture

surfaces of both alloys are complex and somewhat difficult to interpret. In several regions,

well-defined cleavage planes are not apparent (see Figure 4.19). These regions appear to

be heavily deformed suggesting the presence of plastic deformation at or near the crack tip.

In Figure 4.19a, a region from the fracture surface of NiAl + Fe has apparently failed by a

mechanism more complex than brittle cleavage. In fact, even the broad cleavage facets

visible in Figure 4.18a and b, show a surface roughness uncharacteristic of classic brittle

fracture (Figure 4.19b).

1 A. J. Duncan and C.T. Liu, unpublished research, 1992.

4.2.4 Microhardness

Microhardness measurements were conducted on the matrix of alloys surveyed by the

ALCHEMI technique and are presented in Table 4.8. Three heat treatment schedules were

used to characterize this matrix and are listed below. The initial heat treatment (Heat A)

was implemented to homogenize the microstructure and minimize thermal defects (i.e.,

triple defects). Figure 4.20a-d contains optical micrographs of four alloys following this

heat treatment. Heat B was a modification to Heat A and was used to (1) develop an idea

of the concentration of thermal defects present at 1073 K and (2) check for precipitation in

alloys containing higher levels of iron. Finally, some material was also quenched from

1573 K in order to determine the combined effects of thermal defects and iron additions.

Upon examination, some preliminary observations are apparent. First, the smallest

increase in hardness with increasing iron is obtained when iron is substituted for both

nickel and aluminum (Ni/Al = 1). There is also little difference between samples furnace

cooled and water quenched from 1073 K indicating that thermal defects and/or second

phase precipitation do not make a significant contribution to hardness for these particular

treatments. Quenching from 1573 K (Heat C), on the other hand, produced a notable

hardening effect when iron was substituted for nickel or both nickel and aluminum but did

not increase the hardness as drastically when substituted for aluminum.

4.2.5 Solid Solution Hardening

The microhardness data were examined in order to consider mechanisms by which iron

hardens NiAl. Figure 4.21 a, b and c are representations of the effects of solute level on

microhardness for the three different substitution schemes. In Figure 4.21 a, the hardening

relationship is illustrated for iron when it is substituted for nickel. As is evident, this

relationship is discontinuous and does not appear to adhere to a linear (t = Kc) or a

parabolic (T = K-Vc) relationship [59]. Quenching from 1573 K adds to this hardening

relationship consistently over most of the solute concentration range. However, the Ni-

Figure 4.16. SEM micrographs of the fracture surfaces from bend test specimens of
NiA1-0.125Fe (a), NiAl-0.25Fe (b), NiA1-0.5Fe (c) and NiAl-IFe (d).

Figure 4.16. (continued)