You are on page 1of 31

Oxidation of Metals, Vol. 44, Nos.

1/2, 1995

A Critique of Internal Oxidation in Alloys During the


Post-Wagner Era
D. L. Douglass*

Received September 6, 1994; revised January 9, 1995

Wagner's classical treatment of internal oxidation (generic name allowing for


reaction with oxygen, nitrogen, carbon or sulfur) assumed ideal conditions
such as uninhibited dissolution of the gas, formation of spherical particles,
diffusion of the oxidant in the solvent as~ the rate-controlling step, equilibrium
conditions, etc. However, during the 455years since his treatment, many obser-
vations have been made to complic~ie the idealized situation suggested by
Wagner. This paper examines the most important modifications with respect
to Wagner's original analysis. The following items are discussed. (a) The role
of solute concentration: The parabolic kinetics are much higher than expected
for Ni-Al alloys due to rapid interfacial diffusion of oxygen along the interfaces
between cylindrical rods of A1203 (perpendicular to the surface) and the
matrix. (b) Precipitate morphology: Spherical precipitates seem almost to be
the exception. A wide variety of forms have been observed, including Widman-
stiitten platelets, cylindrical rods, hexagonal plates, dendritic or 'fishbone"
products, etc. The competition between nucleation and growth is useful to
explain the observed structures. (c) Intergranular internal oxidation: Rapid
oxygen diffusion in grain boundaries may lead to a wide variety of inter-
granular-precipitate structures. (d) Internal-oxide bands." Wavy, approxi-
mately parallel bands form at a finite distance beneath the surface in certain
alloys having very reactive solutes, e.g., Ag-Mg. It is postulated that high
stresses generated by precipitation play a major role. (e) Surface nodules of
pure solvent metal." High stresses generated during precipitation cause extru-
sion of solute through dislocation pipes, leading to extensive nodule formation
on either grain boundaries or on the grains (or both), depending on the alloy
and oxidizing conditions. (f) Nonstoichiometric precipitates: Either hypo- or
hyperstoichiometric particles can form as very small clusters in certain alloys

*Arizona Materials Laboratories, University of Arizona, Tucson, Arizona 85712.


81
0030-770X/95/0800-0081507.50/0 9 1995 Plenum Publishing Corporation
82 Douglass

(Ag-Al). The nature of precursors and changes in stoichiometry during


reaction are discussed. (g) Trapping of oxidant: Diffusion of the oxidant may
be slowed appreciably by trapping with the solute, although no precipitates
need to form. Lower-than-expected kinetics (based on normal diffusivities of
the oxidant) result. (h) High-solubility-product precipitates: Concentration
profiles of solute, oxidant andprecipitate are quite different than those expected
for low-solubility-product precipitates as considered by Wagner. In particular,
a variable mole fraction of precipitate exists, and further precipitation occurs
in the reaction zone after the front has passed by. Linear kinetics have been
observed for some Nb-base alloys at very high temperatures and low oxygen
pressures. The rate-controlling step is the arrival of oxygen at the surface and
not oxygen diffusion in the metal. (i) Dual oxidants: Two gases may diffuse.
simultaneously and each forms its own product with the solute. The thermo-
dynamically most-stable compound forms near the surface, and the less-stable
compound deeper in the alloy. The less-stable compound is subsequently con-
verted to the more-stable compound with a concomitant release of the second
oxidant. Although numerous examples have been reported of systems which
do not behave as predicted by Wagner, his theory still remains as the corner-
stone of our understanding and is still the starting point for virtually every
study in internal oxidation.

KEY WORDS: internal oxidation; wagner analysis; non-ideal behavior.

INTRODUCTION
Although the phenomenon of internal oxidation had been known for many
years, it was not until 1959 that it was thoroughly analyzed and formalized
by C. Wagner) (Internal oxidation is used in the generic sense in which a
diffusing species from the surface reacts with a less-noble solute in the alloy
to form discrete particles. The species usually causing the phenomenon are
oxygen, nitrogen, hydrogen, sulfur and carbon. Halogens can also cause
internal "oxidation," but the number of cases is very limited.) Wagner m a d e
several assumptions, which at the time, seemed quite reasonable. These
included the following: stoichiometric oxide precipitated at the reaction
front, the formation of low-solubility-product oxides caused virtually all
of the reactive solute to precipitate leaving a matrix of "pure" solvent,
precipitation was complete at the reaction front and no further precipitation
occurred within the internal-oxidation zone subsequently, uniform planar
fronts existed, a constant mole fraction of solute oxide existed across the
reaction zone, and thickening of the front was controlled primarily by oxy-
gen diffusion in the substrate. Wagner's work led to several mathematical
relations that have since been reviewed by numerous authors including
Kofstad,2 Rapp, 3 Swisher, 4 Birks and Meier, 5 among others. The most exten-
sive review was published in 1974 by Meijering. 6
A Critique of Internal Oxidation in Alloys 83

Considerable interest in the phenomenon has persisted over the years


and has led to many publications. There have been many notable exceptions
to the ideas proposed by Wagner, some of which were summarized by
Douglass 7 in 1991. The present paper concerns some, but not all, of the
more pertinent observations over the past 45 years. They will be presented
briefly and compared to the ideas given b y Wagner. The paper is not an
exhaustive review of the field, but rather it is a critical analysis highlighting
significant trends and observations.
Perhaps the most significant factor to emerge is the classification of two
types of behavior. First, there is Wagnerian behavior involving oxides having
very low solubility products. Second, Laflamme and Morral 8 considered
high-solubility-product oxides for which the behavior is quite different. In
fact, Laflamme and Morral formulated a criterion, Eq. (1), showing that
internal oxidation occurs for values falling between zero and unity:

0 < K, s p // ' 'M(s)


O 1,~,r(o)/-
B "" 1 (1)

where Ksp = solubility product of oxide


N(os) -- oxygen solubility at the surface, mole fraction
N(B~ = mole fraction of solute in the original alloy

Wagner's considerations fell within the low end of the criterion, whereas
Laftamme and Morral's work involves the high end of the criterion. The
latter involves relatively few systems and cases compared to the former,
because the oxygen, nitrogen, etc., solubilities in most metals are very low.
Exceptions involve Group IV and V metals that form very stable oxides,
and thus the available solutes that would preferentially oxidize are very
few. For purposes of expediency, the Wagnerian behavior (low solubility
products) will be deemed Type I, and the behavior involving oxides having
high-solubility products is referred to as Type II. Most of this paper involve
Type-I systems.

THE ROLE OF SOLUTE CONCENTRATION


Wagner's analysis yields the following relation, which pertains to one
limiting case:

( 2N(~D~t) '/2 (2)


~=\ vN(BO>
84 Douglass

~I' I I I
]~ .1100~ ~or lOhr
[~ o Ni - AI
]~ r-1 Ni -Cr
I| 9 Ni -Mn
2OO~-~--'
" X Ni-Mo -
I I I

II ,, Ni-V
lain ] I 9 Ni-W
I I
~ o

01 ] J w
0 0-025 0.05 0.075
Atomic fraction solute

Fig. 1. Depth of the internal-oxidation zone as a func-


tion of solute content for various dilute nickel-base
alloys exposed for 10 h at 1100~ in a N i - N i O packfl

where ~ = internal-oxidation zone thickness


Do = diffusivity of oxygen in the substrate
v -- stoichiometry parameter in the oxide
t = time

The relation states that the zone thickness varies inversely with the square
root of the solute concentration. Thus, for a given solvent, e.g., nickel, if
the surface oxygen solubility and oxygen diffusivity are independent of solute
concentration (generally likely for dilute solutions), it should make no
difference what the solute is. In other words, a given amount of solute
should produce the same thickness of the internal-oxidation zone. Stott et
al. 9 studied a number of dilute nickel alloys and generally found that Wag-
ner's prediction was indeed valid. However, one notable exception was
abundantly evident. Ni-A1 alloys "marched to their own drummer," as can
be seen in Fig. 1, which is a plot of zone thickness versus atomic fraction
of solute. All alloys produced the same thickness for a particular solute
concentration except for A1 which clearly enabled much thicker zones to
form than predicted by Eq. (2). Stott et al. attributed this behavior to the
A Critique of Internal Oxidation in AHoys 85

formation of rodlike precipitates of A1203 which were perpendicular to the


alloy surface. The effective oxygen flux, Jeff contains three components:
Jeff= J1Al+ JiAi + JoxAox (3)
where Jr= flux through metal
Z = flux through interface
Jox = flux through oxide
and the respective fractional areas of A1, A~ and Aox. The flux through the
oxide is assumed to be negligible, thus an effective oxygen diffusivity can be
written as
Derf= DoA1 + DoiAi (4)
For cylindrical rods of radius r, the number of rods in unit cross section of
the zone, Z, is given by:
NBO,~,/ Fall= Z Tv2 ~/ Fox (5)
where Van= molar volume of the alloy
Vox= molar volume of oxide
NBov= mole fraction of oxide relative to mole fraction of solute
in the zone
The interfacial area is given by 2rcrZ6~, where 5~ is the interface width,
and Al is given by
&=1 - Ai - Aox = 1 - 2JrrZ6i - Zzrr 2 (6)
Equation (4) can be rewritten

Doe~=Do[1-2/rr3iNB~176 /rr 2 Na~176


:/'Fr2Fall /'EV2Fall J

+D~ [2rcr~i NB~ V~215] (7)


Do [ ~r2Vall J
or

Do Vall --~-- \ D o l -- 1 (8)

Because Doi is much greater than Do, Eq. (8) approximates to

Doe~_14 NBovVox [ 2~iDoi ]


Do Fall L r--Doo 1 (9)
Equation (9) shows that the effective diffusion coefficient of oxygen in
the internal-oxidation zone is a linear function of the density of oxide rods.
86 Douglass

900~

l..1612

"T,.,,, 3 ]612 0
~E

o/
~o
Q

Z
8 2.)512

j612i o
/
J
o/
/
I
0 ~02 0.J0Z. O.J06 J
0-08
NAt
Fig. 2. Plot of permeability versus aluminum content in
Ni-A1 alloys internally oxidizedat 900~ in a Ni-NiO pack
showing the validity of Eq. (10) for the enhanced diffusivity
of oxygen in the interfaces between A1203 rods and the
matrix.

Equation (9) can be expressed also as the product of N(~)Dooee

l,~r(s)n -N(oS)Do 1 + - Vau


o "-'oo~- - / 1 (10)

Since NBov is approximately the same as NB, a plot of N(oS)Dooff versus NB


should give a straight line. Stott et al. showed that this did indeed occur,
(see Fig. 2) for Ni-A1 alloys at several temperatures.
The above treatment of the unexpected behavior of Ni-A1 alloys com-
pared to other dilute Ni-base alloys clearly shows that the precipitate mor-
phology can play a major role in the kinetics of zone growth.

PRECIPITATE MORPHOLOGY

Three facts become readily apparent upon consideration of precipitate


morphologies. First, contrary to Wagner's assumption of spherical particles,
these seem to be the exception rather than the rule. In fact, it is very difficult
A Critique of Internal Oxidation in Alloys 87

8000X
Fig. 3. Dendritic morphology of SiO2 formed during inter-
nal oxidation of Cu-Si near the reaction front. The den-
drites subsequently broke down into "spheres. ''1~

to find cases involving spherical precipitates. Second, there seems to be an


unlimited variety of morphologies to the extent that each alloy is almost
unique. Third, there is generally a wide variation in morphology within the
zone for a given alloy with respect to size, shape, and number. There even
appears to be differences from grain to grain within an alloy. Some examples
of the various types will be given, although not exhaustive, and some com-
ments about the factors controlling morphology will be made.
B6hm and Kahlweit 1~ considered the competition between nucleation
and growth of precipitates at the reaction front, predicting that the number
density of spherical particles varies inversely with the distance from the
surface cubed. In other words, nucleation dominates initially near the sur-
face, and numerous particles exist, whereas growth dominates subsequently,
and the number of particles per unit volume decreases but their size increases
(the mole fraction of oxide across the zone is constant).
Bolsaitis and Kahlweit 11 selected the Cu-Si system to study the size
distributions of very "spherical" particles of SiO2 formed during internal
oxidation. Surprisingly, it was observed that dendrites formed initially, (Fig.
3), and the dendrites subsequently broke up to form fine spheres. The den-
drites formed at the reaction front, but later the side branches disappeared
until eventually spheres existed.
Widmanst/itten platelets form rather commonly in a wide variety of
alloys, over a wide range of conditions, and in various gases. Figure 4 shows
examples of H f N in Nb-30,12 NfO2 in Nb-10Hf, 13 A1N in Ni-10Cr-5A1,14
ZrO2 in Nb-2.5ZrJ 5 The platelets tend to form away from the surface rather
88 Douglass

(a)

200X

~;i 84
iiiii~

Fig. 4. Examples of Widmanst/itten precipitates formed during


internal oxidation and internal nitridation. (a) HfN folaned in
Nb-30Hf at 1700~ in 1 torr N2, 33 rain, ~2 (b) HfO2 formed
in Nb-10Hf at 1768~ in 5x 10-4tort, 10min, 13 (c) A1N
formed in Ni-10Cr-5A1 a t 900~ 14 (d) Z r O 2 formed in Nb-
2.5Zr at 1768~ in 5 x 10 -4 torr, 10 min. 15

t h a n n e a r the surface. Plates a n d needles seem to be u n s t a b l e in some cases,


b r e a k i n g up into smaller g l o b u l a r particles. F i g u r e 5 shows W i d m a n s t f i t t e n
plates o f HfO2 in N b - 1 0 H f b r e a k i n g up 13 a n d TiO2 a n d / o r CoTiO3 needles
in C o - 3 . 7 T i " s p h e r o i d i z i n g . ''~6
M e g u s a r a n d M e i e r 46 p r o p o s e t h a t needles ( a n d p r o b a b l y W i d m a n -
st~.tten platelets also) f o r m b e c a u s e the g r o w t h velocity at the r e a c t i o n f r o n t
A Critique of Internal Oxidation in Alloys 89

(e)

(d)

500X
Fig. 4. Continued.

is essentially equal to the velocity of the advancing reaction front, and thus
the solute has time to diffuse to the tips of the needles and plates. The reduced
rate of reaction-front movement enables growth processes to dominate over
nucleation.
The factors governing the shapes of precipitates have been addressed
by many authors and nicely summarized by Fine. a7 In essence, high interfa-
cial energy between a precipitant and matrix favors a spherical geometry to
minimize interfacial area and thus the total energy, but high strain energy
is minimized if disks, needles, or plates form. The initial precipitation is
most likely determined by minimization of diffusion distances, e.g., pearlite
90 Dou~a~

(a)

(b)

Fig. 5. Examplesof breakup ("spheroidization") of (a) Wid-


manst~tten platelets (HfO2 in Nb-10Hf)13 and (b) TiO2
needles (TiO2 in Co-3.7Ti),~6 after the reaction front has
passed.

formation during transformation of austenite in Fe-C alloys, but the final


shape is determined by energy factors as evidenced by the spheroidization
of Fe3C plates in pearlite. Unfortunately, data are not available for the
various interfacial energies, and thus based on circumstantial evidence
(breakdown a n d / o r spberoidization) of needles and platelets, it appears
that the interfacial energies in those systems are high, and the equilibrium
precipitate shape tends towards a sphere.
A Critique of Internal Oxidation in Alloys 91

Fig. 6. "Fishbone" morphology,TiN formed in Co-3Ti at 1100~ in 49 h.


The sample was deep etched to reveal the structure.

Birks and Meier 5 summarized the many factors governing the size of
the particles. They note that large particles are favored by high particle-
matrix interfacial energies. Megusar and Meier 16 found particles two orders
of magnitude larger in Co-Ti alloys compared to particles in Cu-Ti alloys
in work by Wood e t al. 18 Wood e t al. were able to measure the dihedral
angle of a grain-boundary particle in a T E M micrograph from which they
obtained a value of 400 700 ergs/cm 2 for the TiO2-Cu interfacial energy.
They attributed the small particles to the low energy. The larger particles
of TiO2 in C o - T i alloys 16 and the breakup of needles formed at the reaction
front are consistent with a much higher interfacial energy between TiO2 and
Co than between TiO2 and Cu.
Another interesting morphology noted by Chen and Douglass ~9 is a
"fishbone," (see Fig. 6), of T I N formed during internal nitriding of Co-3Ti.
This structure seems to be intermediate to needles and dendrites and can be
best described as needles with small sidearms. Breakup of the fishbones
occurred also, particularly at higher temperatures, 1100~

I N T E R G R A N U L A R INTERNAL OXIDATION
Grain boundaries are sites for heterogeneous nucleation and act as
short-circuit or rapid diffusion paths. Thus, it is not surprising that
92 Douglass

c)
Fig. 7. Various types of internal-oxidationstructures, after Schenk.2~(a) oxygendiffusiononly
in grain boundaries, (b) oxygendiffusionthrough the alloy lattice, (c) oxygendiffusionmainly
by grain boundaries, (d) diffusion in both lattice and grain boundaries with lattice diffusion
predominating.

precipitation should occur in grain boundaries during internal oxidation.


Schenck e t al. 2~ considered the nature of the reaction front and suggested
that four limiting types exist, as shown in Fig. 7. These are based on
the relative diffusion rates of oxygen as noted in the figure caption.
However, it appears that there may be other types, or perhaps some of
the observed structures are combinations of two or more of the Schenck
types. Figure 8 shows extensive intragranular precipitates as well as larger
grain-boundary precipitates with denuded zones at the boundaries. 15 This
structure is quite similar to those found in age-hardened aluminum alloys.
A combination of Schenck's Types I and III is shown in Fig. 9 for Fe-
51.9 a t % / C r sulfidized at 800~ 21 A network of chromium sulfide exists in
the grain boundaries, but a finite thickness of sulfide particles has formed
within the grains by lattice diffusion of sulfur.
Shida et aI. 22 made extensive studies of internal oxidation in Ni A1
alloys in both Rhines packs (Ni/NiO) and in air. Intergranular oxidation
occurred in the Rhines packs over the range of 800 to 1100~ and was
most prevalent at the lower temperatures. The resulting structures were very
complex and are shown schematically in Fig. 10. Intergranular oxidation was
attributed to stress development during the oxidation process. The stresses
facilitated nucleation and continuous development of oxide in the grain
boundaries. The stress-induced intergranular oxidation was less prevalent at
higher temperatures due to stress-relief processes such as grain-boundary
sliding or general plastic deformation. The less-prevalent intergranular oxi-
dation in air was attributed to injection of vacancies from the growing scale
at the NiO-substrate interface. The vacancies accommodated the volume
increase caused by internal oxidation.
A Critique of Internal Oxidation in Alloys 93

(a)

lO00X

(b)

500X
Fig. 8. Intragranular precipitates, grain-boundary precipitates,
and denuded zones adjacent to the boundaries.~5 (a) Nb 1Zr
oxidized 5 min at 1555~ 5 x 1 0 - 4 torr 02, (b) Nb-2.5Zr oxid-
ized 20rain at 1555~ 5 x 10-4 tort 02.

INTERNAL OXIDE BANDS

Internal oxidation of silver-base and copper-base alloys containing very


reactive solutes such as Mg, Si, Be and A1 m a y cause the formation of
internal bands of oxide that are wavy and roughly parallel to the surface,
as seen in Fig. l lc, d. There appears to be a precipitate-free zone at the
surface, and the exterior region of the internal bands is separated from the
surface by some finite distance. The precipitate-free zone (PFZ) is very hard,
as is the matrix between the bands, although the matrix between bands is
softer than the PFZ. 23 The bands form at higher rather than lower tempera-
tures, e.g., A g - M g alloys did not form bands in 24 h at either 400 or 500~
and exhibited very straight planar interface between the internal-oxidation
94 Douglass

10{] um
Fig. 9. Combined grain-boundary and intragranular inter-
nal sulfidation (Schenck et al. types I and III). Fe-51.9 a/o
Cr sulfidized 72 ks, 1073 K, Ps2 = 10 Pa.

zone and matrix. Planar interfaces formed initially (10 min) at 600~ but
after 1 h internal bands formed, and the reaction front was very irregular
(Fig. 1la, b). Thickening of the bands followed the parabolic rate law with
an activation energy of 17 kcal/mol. This value is higher than that for oxygen
diffusion in pure silver (11.0 kcal/mol.), thus it appears that oxygen diffusion
in the alloy is slower than in pure silver. A significant observation was that
numerous nodules of pure silver formed on the surface (a phenomenon to
be discussed in a separate section) which is indicative of high stresses caused
by precipitation.
The existence of internal-oxide bands has been known for many years
and has been extensively noted in the literature, but there is no adequate
explanation that answers all questions regarding this behavior. Various pos-
sibilities have been discussed by Douglass, 7 who noted the limitations of
each proposed mechanism. The most viable mechanism relates to stress
generation as proposed by Meijering. 6'24'25 The mechanism is analogous to
the formation of healing layers in binary alloys at the base of the solvent-
metal oxide after steady-state conditions are achieved. Enhanced diffusion
to the surface occurs via dislocations generated by the internal-oxidation
A Critique of Internal Oxidation in Alloys 95

Ty_p~. a Ty.pe b
surface
9 o I 9 o e~ f e *a

'. , ,11, '~r ,I'i',


li'I;Ow,,~ If,,,~Ipl111

~1~
l t
grain boundary grain boundary

Type c Ty.pe d
surface

1,
I
i

't l
ub- g rain
I
boundary
grain boundary
gram boundary
Fig. 10. Schematicstructures of four types of intergranular oxidationin dilute
Ni-AI alloys oxidizedin a Ni-NiO Rhines pack. (Shida et al) 2)

process which causes large stresses to exist. Dislocation diffusion provides


solute to the bands. The bands are rather discontinuous and do not provide
protection against further oxidation.

SURFACE N O D U L E S OF P U R E - S O L V E N T M E T A L
As noted in the previous section, nodules of pure solvent form on the
surface of certain alloys undergoing internal oxidation. Guruswamy et al. 26
investigated the nodule formation both with and without imposed strain,
e.g., simultaneous creep. Their interpretation is perhaps the most viable for
the nodule formation and will be considered subsequently.
96 Douglass

(a) 2 hr 500 ~

(b) I0 mln 600~

(C) 65 rain 700~


Fig. 11. Examples of planar reaction fronts at (a)
low temperatures (500~ or (b) short times at
higher temperatures (600~ and internal-oxide
bands formed at (c) 700 and (d) 800~ in Ag-3Mg
oxidized in air. (Douglass et al. 23 )
A Critique of Internal Oxidation in Alloys 97

(d) 60 m• 800~
Fig. 11, Continued.

Guruswamy e t al. observed that the total nodule volume was equal to
the volume change caused by internal oxidation. Nodules formed over grain
boundaries in the grip region of creep samples for some alloys under certain
conditions, whereas grain boundaries in the gage section were devoid of
nodules. On the other hand, unstressed Ag-3.5In oxidized at 800~ showed
grain boundaries devoid of nodules and partially cracked. The same alloy
at 700~ and unstressed showed preferential nodule growth over grain boun-
daries. These examples are shown in Fig. 12.
Douglass e t al. 23 always found preferential nodule formation over grain
boundaries of Ag-3Mg and an increase of nodules with time, Fig. 13. Rubly
and Douglass 14 also found nodules of pure nickel on various Ni-Cr-A1
alloys that had been internally nitrided, although there did not appear to
be any difference in nodule formation over grain boundaries compared to
the grains. Nickel nodules in plan view and in cross section are shown in
Fig. 14.
Guruswamy and co-workers considered various models to explain nod-
ule formation. The Nabarro-Herring creep mechanism was assumed, but
calculated upper and lower bounds for rate constants for internal oxidation
based on diffusional flow were much lower than observed values. It was
further found that the large volume change caused by internal oxidation
could not be accounted for by vacancy flow. Thus, the region around each
particle forming at the reaction front must deform by slip rather than by
creep, and slip increases the dislocation density. Countercurrent silver/
vacancy flow must therefore occur by dislocation-pipe diffusion, this being
the only mechanism accounting for the large amount of silver transported
to the surface. The slight increase in the internal oxidation rate when samples
were subjected to creep is caused by enhanced generation of dislocations.
98 Douglass

(a)

(b)

Fig. 12. Surfacenodules of pure silver formed on Ag-3.5In after 2 h


oxidation at (a) 800 and (b) 700~ (Guruswamyet al. 26)

NONSTOICHIOMETRIC PRECIPITATES
Certain very reactive solutes in much-more-noble solvents, i.e., A1 in
Ag and Mg in Ag, have been shown to form non-stoichiometric particles
which can be hypo- or hyper-stoichiometric. Darken 27 reported on studies
of Ag-A1 by T E M and low-angle X-ray scattering in which very small clus-
ters of A1 and oxygen formed. The clusters had O:A1 ratios greater than
1.5, the value corresponding to A1203. The variation of the O/A1 ratio with
A Critique of Internal Oxidation in Alloys 99

40000

Cq

E 30000
E

0
20000
Z

<

10000

0 100 200

t (rains.)
Fig. 13. Effect of oxidation time on the number density of silver nodules
formed on the surface during oxidation of Ag-3Mg at 580~ (Douglass
et al. 23 )

temperature as well as the temperature dependence of size are shown in Fig.


15. Hyperstoichiometry can be explained for very small clusters according
to the schematic arrangements in Fig. 16.
Combe and co-workers2s-3~studied internal oxidation in Ag-Mg alloys
using a variety of techniques. They too found O/Mg ratios greater than the
stoichiometric value and further found changes with time, as noted in Fig.
17. They concluded that rapid fixation of oxygen occurred initially followed
subsequently by slower growth by coalescence of the oxidized species. It was
proposed that "elementary species" formed, e.g., MgO2, corresponding to
a value much greater than the stoichiometric ratio. The decrease in Fig. 17
from the maximum value to (O/Mg)min occurred by oxygen release from
the superstoichiometric cluster. The Mg atoms are initially bound to oxygen
atoms and are inhibited from diffusing. The clusters reorganize slowly, re-
leasing some oxygen. Rearrangement at higher temperatures causes the clus-
ters to become more compact, in agreement with Darken's observation. The
more-weakly-bound oxygen atoms at the periphery of the clusters break
their bonds. They are less strongly bonded, because they have few bonds
with Mg but many with Ag atoms. The unusual behavior of the very reactive
solute in noble solvents may be related to the nonstoichiometry of the parti-
cles and its change with time.
I00 Douglass

(a)

(b)

i;i

Fig. 14. Pure-Ni surface nodules formed during internal nitriding


of (a) Ni-20Cr-5A1 nitrided 24 h at 800~ plan view, and (b)
Ni 5A1 nitrided 48 h at 900~ cross section. (Rubly and
Douglass.14)

There has been no evidence to date of nonstoichiometric oxides forming


during internal oxidation of the less reactive solutes. Wagner's assumption
of stoichiometric particles forming is valid for most systems, but as noted
above, there are some significant exceptions.

T R A P P I N G OF O X I D A N T
The transition from internal to external "oxidation" in Ni C~c alloys
requires a much higher chromium content when nitrogen is the oxidant
compared to oxygen. The approximate values, depending upon temperature
and pressure, are 30 to 40 at.% for CrN-film formation and about 15 at.%
A Critique of Internal Oxidation in Alloys lOl

TEMPERATURE, " F
600 800 I000 1200 1400 1600
5.0 - 1 - , , , ,

2.5

o
5
~: z.o t5
d
20
_AI~O 3 . . . . . . . . . . . . . . . . . . 25
i.5
I I
500 ,| ,;o 8oo 9OO
OXIDATION TEMPERATURE, *C

Fig, 15. The O/AI ratio in A1-O clusters formed during internal oxida-
tion of Ag-AI alloys as a function of temperature (left scale) and the
cluster size (right scale). (Darken. 27)

0 0 0 0 0 0 0
A~--A~ AI-A~ A~
o o o o o olololo
AI '" AI AI AI AI AI
0 0 0 0 0 Ol O l 0 I 0
AI AI AI
0 010 0

Fig. 16. Schematic arrangements of aluminum-oxygen clusters on a silver lattice. (Darken. 27)

for Cr203-film formation. 14 This surprising result contradicts the above


observation, because the diffusivity of oxygen compared to nitrogen in nickel
should be more rapid on the basis of atomic radii of 0,70 A for nitrogen
and 0.61 A for oxygen. A slower diffusivity of the oxidant in the solvent
would allow more time for lateral growth of the precipitates at the reaction
front, in which case coalescence of the particles could occur and thus form
a continuous film. Rapid oxidant diffusion favors nucleation of new particles
rather than growth of existing particles. Internal-nitridation rates and perme-
abilities clearly show that a higher critical solute level is needed for the
formation of a continuous CrN film compared to formation of a Cr203 film.
102 Douglass

0
kq

o_

rain
0.5

, Jl it
so'oo ~o'ooo ~5'boo 20000
t(s),

Fig. 17. Variation of O/Mg ratio during internal oxidation


of Ag-0.5Mg at 345~ (Charrin eta/. 3~

Permeabilities, o b t a i n e d f r o m m e a s u r e m e n t s o f the thickness o f


i n t e r n a l - o x i d a t i o n zones, are shown in an A r r h e n i u s p l o t (Fig. 18), 14 a l o n g
with d a t a o b t a i n e d b y P a r k a n d A l s t e t t e r 3~ using a n electrochemical tech-
nique. T h e l a t t e r are consistent with i n t e r n a l - o x i d a t i o n data. Permeabilities

Temperature(~C)
900 800 700
10-1o , , ,
=.,,.-.,---~ Ni-20Cr.-5AI(int. Nitrd.)
~ Ni-20Cr (Int. Nitrd.)
~" Ni-10Cr
Eo ~ "~ "~ (Int. Nitrd,)

:z~ 10-~2
d Ni-5,,.,,I .. Ni-10C.r-SAI
z= (Int, ~ ' i t r d . ) ~ ~ , . . "~,lnnt' Nmtrd)
10<3 Ni-SAI/ ~,, ~".
9
(int. Oxide) / . ~~\ - , .-., . \ Nl-10Or
NoDo/ ~ " , . (Int. Nitrd,)
10-14
(Park & Alstetter)
I I I I I
8.0 9.0 10,0 11.0
1/-r x lO4 (l/K)
Fig. 18. Arrhenius plot of permeabilities, NNDN and
NoDo, determined from internal nitridation, internal
oxidation, and electrochemical measurements] 4
A Critique of Internal Oxidation in Alloys 103

are always greater for internal nitriding than for internal oxidation with the
exception of Ni-5A1 which, as noted, 9 forms A1203 rods perpendicular to
the surface, and thus rapid diffusion occurred along the alumina/matrix
interfaces. Nitrogen diffusivities in nickel are not known, but results calcula-
ted from measured permeabilities and available nitrogen solubilities 7 clearly
showed that nitrogen diffusivities were about 100 times faster than those of
oxygen. The calculated values were opposite to the expected values based
on size considerations and diffusivities of oxygen and nitrogen in other
metals. 2 The only logical explanation for the observed permeabilities and
calculated diffusivities is that oxygen is trapped at the dislocations or some
other sites and that a high binding energy exists between the oxygen atoms
and the trapping sites. Trapping of hydrogen in iron 32 and of oxygen in
dilute refractory-metal alloys have been reported. 33'34
It is apparent that both internal-oxidation rates and the critical~ values
for transition are markedly affected by subtleties in diffusion behavior,
neither one of which were considered by Wagner.

HIGH-SOLUBILITY-PRODUCT PRECIPITATES
The high end of the Laflamme-Morral criterion, Eq. (1) (Type-II
behavior), represents a very interesting and somewhat unique aspect of inter-
nal oxidation. Type-II behavior occurs in systems in which appreciable solu-
bility of oxygen exists as seen schematically in Fig. 19.* The variation of
solute concentration, oxidant concentration and mole fraction of precipitates
across the reaction zone as well as in the substrate was calculated by Christ et
al., 35 who used a finite-element analysis to analyze TiC precipitation during
internal carburization in Ni-Ti. Their results, replotted, are shown in Fig.
20. Several features are readily apparent that should be compared to Type-
I behavior, for which plots are given by Rapp, 3 Birks and Meier, 5 etc. First,
the solute mole fraction does not go to zero in the internal-reaction zone
but decreases from the reaction front to the surface. Second, the oxidant
activity does not go to zero at the reaction front but decreases monotonically
into the alloy beyond the reaction front. Third, the mole fraction of precipi-
tates varies across the zone, and further precipitation can occur even after
the front has passed by.
The distribution of precipitates was considered and analysed by Ohriner
and Morral. 36 Their calculated results are shown in Fig. 21 as a normalized
set of distribution curves for several values of a solubility parameter, a
(described on the figure). The value of unity for alpha is for Type-I behavior,
whereas alpha approaching zero is the other extreme for an initially saturated
alloy for which the precipitation fraction equals erfc(y). Values of alpha
between zero and one represent various degrees of Type-II behavior.
104 Douglass

~ X ,,-
BC 9 Moles of BC i
nmo x

Cs ~~~nc. ofsoluble
IC
I

I
I

. . . . J ~
0
X ~
(ol

Umif

/ o X~/~.OCrOSS subscole

*/.B ,,-
(b)
Fig. 19. Alloy system exhibiting Type-II behavior of internal oxida-
tion, based on the Laflamme Morral criterion. 8

0.08
0.06 ../(NB)'~

o~.
z
0.04
(o)~,~
~" 0.02 ~ 0 ) ~

0.0
0.0 2.0 4.0 6.0
DistancefromSurface(mm)
Fig. 20. Calculated profiles for high-solubility-product
oxide formed during internal oxidation. (Christ
et aL 35 )
A Critique of Internal Oxidation in Alloys 105

1.0

.8

~
0 .6

.4

a,,)

.2

. ?. .4 .6 .8 1.0 h2 1.4 1.6 1.8 2.0

y = X

2~/D+N(~) ~,~'r(BO')n'IT~(BO~)B
0

o~ = m a x . v a l u e o f solute in precipitate
= 1 - Ksp/N~)- N(~)
Fig. 21. Normalized distribution curves of precipitates for various values of a
solubility parameter. (After Ohriner and Morral. 36)

A combination of very high temperature (rapid oxygen diffusion in the


substrate) and low-oxidant pressure (limited supply and arrival rate) cause
a transition in kinetics from the parabolic rate law to a linear rate law, 12'13'15
as shown in Fig. 22. Linear kinetics are clearly due to the slow rate of oxygen
arrival at the metal surface. In order to test this hypothesis, Corn e t al. 13
used the kinetic theory of gases, assuming a sticking coefficient of unity for
the gas, to calculate a maximum rate. The maximum rate based on this
106 Douglass

70,
KINETIC THEORY , 1768"C 1555"C

./ /

lIE

10

1000 2000 3000

TIME, SECONDS

Fig. 22. Linear kinetics of the internal oxidation of C129Y ( N b -


10Hf-0.2Zr-0.1Y) at 5 • 10 -4 torr oxygen. ~3

calculation was only slightly higher than the measured rate of a N b - H f


alloy, C129Y, as shown in Fig. 22. The conclusion that the rate of gas arrival
was indeed the rate-controlling step was further substantiated by using argon
containing 1.3 ppm oxygen, giving an equivalent oxygen partial pressure of
10 - 4 torr used in the other experiments. No internal oxidation occurred in
the argon containing oxygen, whereas internal oxidation occurred in pure
oxygen at 10 - 4 torr. This indicates that the arrival of oxygen at the metal
surface was further inhibited by the argon atoms which blocked the oxygen
molecules and effectively prevented them from striking the metal.
On the other hand, experiment at lower temperatures in Rhine's packs 3v
resulted in parabolic kinetics for the same alloys. Even though the oxygen
pressure was considerably less in the Rhine's packs, oxygen diffusion in the
substrate was orders of magnitude slower at the lower temperatures, and
thus oxygen arrival at the surface was relatively rapid compared to oxygen
diffusion. Examples of the parabolic kinetics are shown in Fig. 23.
A Critique of Internal Oxidation in Alloys 107

250-- ='" -/--"\1


-' 1060 o C

~[ ~9 C6 0/ ~ ~ ~ A
A 200
E: I
"1-
I-
Lu
I1.
150-
13
I-.,
z
n~
0
u.
zo_ 100-

I-
0

IZ
ILl
50

' I ' I ' I


0.00 4.00 8.00 12.00
TIME (hours)
Fig. 23. Kinetics of internal oxidation of CI03 (Nb 10Hf-0.5Zr-0.STi) showing
parabolic behavior at low temperatures. Oxidation performed in a Rhines pack of
Ni-NiO. 37

The rate-controlling step has been assumed, almost universally, to be


oxygen diffusion in the metal, and generally this is true. However, if the
temperature is sufficiently high so that oxygen diffusion is very rapid, then
transport of oxygen in the gas phase becomes the rate-controlling step. Most
systems studied with respect to internal oxidation involve base metals whose
melting points are quite low compared to Nb, thus oxygen diffusion in the
substrate is very low, and therefore gas-phase diffusion is not a consideration
in those systems.

DUAL OXIDANTS
There are many cases in which dual oxidants exist, such as air which
has been used over the years for oxidation studies. Generally, most metals
do not react with nitrogen, although there are some notable exceptions such
as Cr, Ti, Zr, Hf, and some Group-V metals. Reaction of a metal with CO
108 Douglass

can cause both internal oxidation and internal carburization. Carbonitriding


is another example of dual oxidants reacting to form internal compounds
from each oxidant. Finally, hot corrosion involves both oxygen and sulfur,
both of which can react internally. In principle, there are many combinations
of gases that can cause internal reaction, i.e., both reactants form their
respective compounds.
Meijering 6 was apparently the first to consider this phenomenon. He
notes that generally oxides are more stable than the compounds formed
from the other gases. Thus, if an alloy is initially internally oxidized and
subsequently internally carburized, etc., the more stable oxide is not reduced
by the other gas, and no compounds of the second gas form within the
internal-oxidation zone. The second oxidant forming less stable compounds
than oxides, can cause reaction only beyond the interface between the alloy
and the internal-oxidation zone. On the other hand, if the alloy is first
carburized and then oxidized, the internal carbides will be oxidized thereby
liberating carbon (or any other oxidant, depending on the gas used initially).
The liberated carbon can either diffuse inward in which case carbides will
form beyond the internal-oxidation zone, or the liberated carbon can diffuse
out of the sample altogether at the surface, depending upon the oxygen
activity. A completely internally carburized flat strip of metal which is subse-
quently oxidized behaves as though the material had never been carburized.
In other words, the carbides are simply converted in situ to oxides. All of
the above assumes that the displacement reaction in which carbide is con-
verted to oxide involves small particles and that the conversion reaction is
rapid. This may not be the case for large particles for which diffusion in
either the carbide or oxide could become rate controlling.
Let us now consider the treatment of Meijering 6 for simultaneous inter-
nal reaction by two oxidants, using the schematic diagram shown in Fig.
24. 24 The thickness of the outer layer, 0, the zone of internal oxides, i.e.,
the more stable compound, is given by
rI = (2NoDot/n2NB2) 1/2 (1 1)
where No is the solubility of oxygen in the solvent, Do is the diffusivity of
oxygen in the solvent, t is the time, n2 is the O/B ratio of the oxide, and NB
is the mole fraction of B (solute) in the alloy. Carbon atoms (or atoms
of any other second oxidant) find unreacted B only beyond 77- Ux is the
concentration of free second oxidant and is given by

Ux = Nx+A1 ere[2 (Dxt)-l/21 for 0 < x < ~ (12)

Ux=A2{erf[~(Dxt)-l/2]-erf[2(Dxt)-'/2]} for r / < x < ( ( 1 3 )


A Critique of Internal Oxidation in Alloys 109

NB
Internal-oxide ] Internal-sulfide (or carbide, etc.) I ~ - - NBo
"#r zone ~!~1 zone ~ /

= Nxs
unreacted
LL
G)
alloy

No s

t n
mixed gas/alloy
interface x--*
Fig. 24. Concentrationprofilesin an alloy for dual oxidants. (Meijering.6)

Nx is the mole fraction of second oxidant in the solvent. There are three
unknowns: A1, A2 and 89 ( ( D A ) -1/2 (hereafter called p). The quantity
89 -1/2 (hereafter called rn) is obtained from Eq. (11). Equations (12)
and (13) must give equal U~ at x - - q ; the difference in the concentration
gradient dU~/dx at both sides of x = q is determined by the amount of the
second oxidant liberated by reaction of oxygen with the second-oxidant
compound. The magnitude of dUx/dx at x = q determines the velocity of
the boundary between the alloy and the second-oxidant reaction zone. Thus,
pe pz erfp = me m2erf m + N~/NB n l zr1/2 (14)
A graph of xe x2 versus eft(x) can be used to solve for p. If No <<NB, N~ <<NB
and NoDo<<N~Dx, both m and p are small, and Eq. (14) becomes
(2 = (2t/NB)(NoDo/n2 + N~D~ In1 ) (15)
This relation states that the square of X penetration driven by oxygen is
increased by the square of the oxygen penetration.

CONCLUSIONS
A critical analysis of the field of internal oxidation during the period
after Wagner first formulated his theory on this subject has been made with
the main goal of analyzing the various observations and theories that are
seemingly contradictory to the idealized assumptions made by Wagner.
Specifically, the following items have been discussed: (a) the role of solute
concentration and its effect on reaction kinetics, (b) precipitate morphology,
110 Douglass

(c) i n t e r g r a n u l a r i n t e r n a l o x i d a t i o n , (d) i n t e r n a l - o x i d e b a n d s , (e) s u r f a c e


n o d u l e s o f p u r e - s o l v e n t m e t a l , (f) n o n s t o i c h i o m e t r i c p r e c i p i t a t e s , (g) t r a p -
p i n g o f o x i d a n t , (h) h i g h - s o l u b i l i t y - p r o d u c t precipitates, a n d (i) d u a l oxi-
d a n t s . E v e n t h o u g h there are n u m e r o u s e x a m p l e s for w h i c h W a g n e r ' s
a s s u m p t i o n s a n d a n a l y s i s do n o t a p p l y , his t h e o r y r e m a i n s as the s t a r t i n g
p o i n t for a n u n d e r s t a n d i n g o f this field.

REFERENCES
1. C. Wagner, Z. Elektrochem. 63, 772 (1959).
2. Per Kofstad, High-Temperature Oxidation of Metals (Wiley, New York, 1966).
3. R. A. Rapp, Corrosion 21, 382 (1965).
4. J. H. Swisher, in Oxidation of Metals and Alloys, D. L. Douglass (ed.) (ASM, Metals Park,
Ohio, 1971), Chapter 12.
5. N. Birks and (3. H. Meier, Introduction to High-Temperature Oxidation o f Metals (Edward
Arnold, London, 1983).
6. J. L. Meijering, Advances in Materials Science, Herbert Herman (ed.) (Wiley, New York,
1971), p. 1.
7. D. L. Douglass, J. Met. Nov., 74 (1991).
8. G. R. Laflamme and J. E. Morral, Acta Met. 26, 1791 (1978).
9. F. H. Stott, G. C. Wood, D. P. Whittle, B. D. Bastow, Y. Shida, and A. Martinez-Villafane,
Solid State Ionics 12, 365 (1984).
10. G. B6hm and M. Kahlweit, Acta Met. 12, 641 (1964).
11. P. Bolasitis and M. Kahlweit, Acta Met. 15, 765 (1967).
12. J. (3. Keller, Ph.D. dissertation, UCLA, 1994.
13. D. L. Corn, D. L. Douglass, and C. A. Smith, Oxid. Met. 35, 139 (1991).
14. R. P. Rubly and D. L. Douglass, High-Temperature Corrosion of Advanced Materials and
Protective Coatings, Y. Saito, B. Onay and T. Maruyama (eds.) (Elsevier, Amsterdam,
1992), p. 133.
15. D. L. Douglass, D. L. Corn, and F. Rizzo, J. Phys. IV, Coll. C9 suppl. J. Phys. III 3, 75
(1993).
16. J. Megusar and G. H. Meier, Met. Trans. A 7A, 1133 (1976).
17. Morris E. Fine, Phase Transformations in Condensed Systems (Macmillan, New York,
1964).
18. S. Wood, D. Adamonis, A. Guha, W. A. Sofia, and G. H. Meier, Met. Trans. A 6A, 1793
(1975).
19. L Chen and D. L. Douglass, Oxid. Met. 38, 189 (1992).
20. H. Schenck, E. Schmidtman, and H. Mfiller, Arch. Eisenhuttenw. 31, 121 (1960).
21. T. Narita, W. W. Smeltzer, and K. Nishida, Oxid. Met. 17, 299 (1982).
22. Y. Shida, F. H. Stott, B. D. Bastow, D. P. Whittle, and G. C. Wood, Oxid. Met. 18, 93
(1982).
23. D. L. Douglass, Baizhao Zhu, and F. Gesmundo, Oxid. Met. 38, 365 (1992).
24. J. L. Meijering, Strength of Solids, Bristol Conference Report (Physical Society, London,
1947), p. 140.
25. J. L. Meijering, Pittsburgh International Conference Surface Reaction (Corrosion, Pitts-
burgh, 1948), p. 101.
26. S. Guruswamy, S. M. Park, J. P. Hirth, and R. A. Rapp, Oxid. Met. 26, 77 (1986).
27. L. S. Darken, Trans. A S M 54, 600 (1961).
28, L. Charrin, A. Combe, and G. Moya, J. Therm. Anal. 14, 89 (1978).
29. A. Combe and J. Cabane, Oxid. Met. 21, 21 (1984).
30. L. Charrin, A. Combe, and J. Cabane, Oxid. Met. 37, 65 (1992).
31. Jong-Wan Park and Carl J. Alstetter, Met. Trans. A 18A, 43 (1987).
32. H. H. Johnson, Met. Trans. A. 19A, 2371 (1988).
A Critique of Internal Oxidation in Alloys 111

33. R. J. Lauf and C. J. Alstetter, Acta Met. 27, 1157 (1979).


34. R. C. Frank, R. J. Lauf, and C. J. Alstetter, Met. Trans. A. 13A, 539 (1982).
35. H. J. Christ, J. H. Biermann, F. Rizzo, and H. G. Sockel, Oxid. Met. 32, 111 (1989).
36. E. K. Ohriner and J. E. Morral, Scripta Met. 13, 7 (1979).
37. Carlos Nunes and D. L. Douglass, unpublished research.

You might also like