Oxidation of Metals, Vol. 41, Nos. 1/2, 1994

Effect of AI on High-Temperature Corrosion of Co-15a/oNb Alloys in H 2 - H 2 0 - H 2 S Environments C. C. Shing* and D. L. Douglass* RecezvedJanuary 14, 1993;revisedOctober 13, 1993

Co-15at.% Nb alloys containing up to 15at.% AI were corroded in gaseous I I 2 - H e O - H 2 S mixtures over the temperature range of 600-900~ The corrosion kinetics followed the parabolic rate law at all temperatures. Corrosion resistance improved with increasing Al content except at 900~ Duplex scales formed on alloys consisting of an outer layer of cobalt sulfide and a heterophasic inner layer. A small amount of Ale03 was found only on C o - 1 5 N b - 15Al. Contrary to what formed in Co-Nb binary alloys, neither NbSe nor NbOe were found in the inner layer of all alloys, but Nb3S 4 did form. The absence of NbS2 and NbOe is due to the formation of stable AleO s and AleS3 that effectively blocked the inward diffusion of oxygen and sulfur, respectively, and to the reduction of activity of Nb by Al additions in the alloys. Intercalation of ions in the empty hexagonal channels of Nb3S 4 is associated with the blockage of the transport of cobalt. An unknown phase (possibly Alo.sNbSe) was detected. Alloys corroded at 900~ were abnormally fast and formed a scale containing CoNb3S 6 and Co. Pt markers were found at the interface between the inner and outer layers. KEY WORDS: corrosion; A1203; Co-Nb-A1; Nb3S4,

INTRODUCTION Metallic materials are generally subject to unusually severe degradation when exposed to gaseous environments at high temperatures. This degradation process consists of the consumption of metal by reaction with the gaseous species of the atmosphere. In past decades, the development of *Department of Materials Science and Engineering, School of Engineering and Applied Science. University of Califorma, Los Angeles, California 90024-1595, 115 0030-770x/94/0200-011550700/09 1994PlenumPubhshmgCorporation

116

Shing and Douglass

corrosion-resistant alloys based on Fe, Ni, and Co for use in high-temperature oxidizing environments has been studied thoroughly. In principal, most of the alloys are designed to resist high-temperature oxidation by adding sufficient amounts of chromium, aluminum, or silicon to form a continuous external scale of their oxides; these oxides act as a barrier that separates the metal and the environment and provide oxidation resistance due to their low growth rates. Nevertheless, there are a number of high-temperature industrial processes, such as oil refining, coal gasification, and fossil-fuel conversion, that consist of complex gaseous environments which contain sulfur-bearing gases (HzS , 502) along with oxidizing species (HzS , CO, CO2). These multireactant atmospheres have very low oxygen activities but appreciable sulfur activities. For example, a coal-gasification atmosphere typically contains partial pressures of oxygen and sulfur of 10-18 atm and 10 -6 atm, respectively, at 850~ 1 The primary corrosion problem in these mixed-gas environments is sulfidation. The sulfidation rate of most high-temperature alloys is generally greater than their oxidation rate by orders of magnitude. This is mainly due to the formation of nonprotective sulfide scales on the alloys. The nonprotectiveness of sulfides is attributed largely to the high degree of nonstoichiometry in sulfides. 2 Compounds with a large deviation from stoichiometry contain a high density of lattice defects and will have high diffusivities. Thus, the high transport rate of matter in sulfides contributes to the higher rate observed in metal sulfidation. Most oxidation-resistant alloys, which rely on the formation of protective Cr203 and A1203 scales, have poor high-temperature sulfidation resistance in sulfurbearing atmospheres. This is mainly because the protective oxide scale either fails to form or the breakdown of the oxide occurs eventually in such environments.3 A number of refractory metals (Nb, Mo, Ta, W) show superior resistance to sulfidation than to oxidation.4 The sulfidation rates of Mo and Nb are comparable to the oxidation rates of Cr. Investigations of the effect of molybdenum and niobium additions on the corrosion resistance of the base metals, such as iron, cobalt, and nickel in H 2 - H 2 0 - H 2 S mixed-gas atmospheres have been studied in this laboratory. 5-1~ The results have clearly demonstrated that superior corrosion resistance is attained. However, the corrosion rate of these binaries was still high compared to that of pure refractory metals. The effect of A1 additions on the corrosion behavior of a Co-15Mo alloy in a gaseous mixture was studied previously and was found to further reduce the corrosion rate below that of the binary alloy. ~ The slow corrosion rate was attributed to the formation of a ternary sulfide, Alo.ssMo2S4, and A1203, which effectively blocked the transport of cobalt ions through the sulfide scale. In addition, a beneficial effect of A1

High-Temperature Corrosion of Co-15a]oNb Alloys

117

was observed also on the sulfidation of F e - 3 0 N b - A 1 alloys. 12 Therefore, it is of interest to study the corrosion behavior of C o - N b - A 1 alloys and to compare it with that of C o - M o - A 1 alloys. The mixed-gas environments used in this study were the same as used previously, 6 with a composition corresponding to Po~ = 10-~~ atm and Ps2 = 10-5 atm at 850~ The gas composition remained constant throughout the experiments regardless of test temperature. EXPERIMENTAL

PROCEDURES

Alloys containing 15 at.% Nb and varying amounts of A1 up to 15 at.%, were studied. The alloy compositions, given in both at.% and wt.%, are listed in Table I. Throughout this paper, alloy compositions will be given in atomic percentage. The constant atomic proportion o f Nb was maintained in order to compare the effect of aluminum additions on the corrosion behavior of C o - N b alloys. In addition, the concentration of Nb was the same as that of Mo in C o - 1 5 M o - A 1 alloys studied previously. 11 Alloys were fabricated into approximately 60-g button ingots by arc-melting under a titanium-gettered atmosphere of high-purity argon. The starting materials, obtained from AESAR Chemical Co., were at least 99.8% pure. During fabrication, the buttons were flipped over and remelted at least five times to ensure homogeneity. The as-cast buttons were sealed in an evacuated quartz tube and then annealed at 1000~ for 72 hr. Specimens approximately 1.0 mm thick were cut with a low-speed diamond-wafering blade. A suspension hole was drilled through each sample by a spark cutter. The weight and surface area of samples were approximately 0.6 g and 2.5 crn ~, respectively. The specimens were then ground using 600-grit SiC abrasive paper, followed by polishing with 6-micron diamond paste. Prior to corrosion, the samples were rinsed in water and degreased in methanol or acetone. The microstructures of the alloys are shown in Fig. 1. C o - 1 5 N b - 5 A 1 is a two-phase alloy, while C o - 1 5 N b - 1 0 A 1 and C o - 1 5 N b - 1 5 A 1 are three-phase alloys. The bright phase in C o - 1 5 N b - 5 A 1 is the niobium-rich Table I. Composition of Ternary Co Nb-A1

Alloys Atomic %

Weight %

Co- 15Nb- 5A1 Co- 15Nb- 10A1 Co 15Nb- 15A1

Co-22.3Nb-2.2AI Co-22.9Nb-4.4AI Co-23.5Nb-6.8AI

118

Shing and Douglass

;i~;.

=

~

o

.6lr~ ,-~

.,

.~ ~. ~.;.

,L

2 1I

.;

'~

"p

1"~_ .

I

,~"

I)

,

9..:-

~, ,,

,

J

I. ~'q

C-"" g, ~ P i.

,.~.

4,0,,

:

.-

L~

* ~

'J"

-"

ar

'i

9 '

_.~

,r--: ,I

..,

;

,; .0 ..,.,,'

*,:

,

I

,~. .2- .~

F

Fig. 1. Microstructures

of Co

,

,

1 5 N b - A 1 alloys: (a) Co 10A1; (c) Co 1 5 N b - 1 5 A 1 .

15Nb-5A1;

(b) C o - 1 5 N b -

intermetallic phase, NbCo3, and the gray phase is a-Co. Co-15Nb-10A1 and Co 15Nb-15A1 contained a third phase, CoAl, which is the dark phase shown in Fig. lb,c. The volume fraction of CoAl increased with increasing A1 content. Table II summarizes the results of the phase analysis using XRD and standardless semiquantitative analysis of EDS. Apparently the two intermetallic phases, NbCo 3 and CoAl, contain very little of the ternary elements, A1 and Nb, respectively. It is worth noting that the

High-Temperature Corrosion of Co-15a/oNb Alloys

119

Table II. Composition of Phases in C o - N b

A1 Alloys

Phase identified by X R D

Co 15Nb-5A1 C o - 1 5 N b - 10A1 C o - 15Nb-15A1

NbCo s

a-Co

CoAl

C070 9Nbas 7A13 4 C069 4Nb25 1Als 5 C~ 2Nb25 ~AIs s

COs8 0Nbz 4A19 6 C087 0Nbl 7Al11.3 C~ 6Nbl 8Al12 6

C059 oNb4 7A1363 C~ 8A[36 s

composition of various phases remains consistent, leading only to changes in the volume fraction of the various phases in all alloys. The volume fraction of each phase in each alloy was determined by using computerized image analysis, Table III. The corrosion experiments were carried out continuously using a quartz-spring thermogravimetric apparatus as described elsewhere. 6 The corrosive environment is provided by H z - H z O - H 2 S gas mixtures, which produce an oxygen pressure of 10 -2~ atm and a sulfur pressure of 10 -5 arm at 850~ Premixed H2-H2S gas (1.9% H2S and 13.4% H2, balance argon) was used. The content of H 2 0 w a s established by passing the premixed gas through water. The corresponding equilibrium partial pressures of 02 and $2 were calculated based on thermodynamic equilibria among various species in the gas mixtures. Temperature-dependent partial pressures are listed in Table IV. Characterizations of the scales on corroded samples were performed by X-ray diffraction (XRD) using Cu-Ke radiation, optical microscopy, Table III. Volume Fraction of Phases in C o - N b - A 1 Alloys Alloys

NbCo s

a-Co

CoAl

Co- 15Nb-5AI C o - 15Nb 10A1 C o - 15Nb 15A1

0.688 0.635 0.597

0.312 0.231 0.132

-0.134 0.365

Table IV. Composition of the Gas Atmospheres at Various Temperatures

Po2 (atm) Ps2 (atm)

600~

700~

800~

900~

2.7 x 10 -27 4.5 x 10 8

2.9 x 10 -24 5.6 x 10 - 7

8.9 x 10 -22 4.5 x 10 - 6

1.0 x I0 -~9 2.5 x 10 - 5

120

Shing

and Douglass

scanning electron microscopy (SEM) equipped with an X-ray energy-dispersive spectrometer (EDX), and electron probe microanalysis (EPMA) using X-ray wavelength-dispersive spectrometry (WDS). Extensive X-ray diffraction analysis was performed on the scales. Initially, the exterior scale was analyzed; the scale was then abraded slightly and examined again. This process was repeated several times until only base-metal peaks were present. The phases present were identified by comparing the patterns obtained with those reported in the Joint Committee on Powder Diffraction Standards (JCPDS) files. Platinum wire, 25 #m in diameter, was used for marker studies. The wire was spot-welded at several points on the cleaned surface of the samples. RESULTS Kinetics

The corrosion of C o - 15 at. % Nb alloys containing 5, 10, and 15 at.% aluminum was studied over the temperature range of 600-900~ in H2H20-H2S mixed-gas environments. Figure 2 shows the parabolic plots of the corrosion kinetics for the three alloys. The parabolic rate law was followed regardless of temperature or alloy composition, although twostage kinetics were noted in some cases. The steady-state parabolic rate constants are listed in Table V. The parabolic kinetics observed for the ternary alloys indicate that their corrosion is controlled by a solid-state diffusion process. An Arrhenius plot, Fig. 3, shows the temperature dependence of the parabolic rate constants. The corrosion rates increased generally with increasing temperature. In addition, the corrosion rates for all the alloys studied were abnormally high at 900~ This behavior was observed also in the corrosion of Co-15Mo-A1 alloys containing up to 10at.% A1.tl In fact, the scale phases and morphology for alloys corroded at 900~ were totally different from those at 600-800~ as mentioned later. Therefore, Table

V. P a r a b o l i c R a t e C o n s t a n t s f o r the C o r r o s i o n o f V a r i o u s A l l o y s in the H 2 - H 2 0 - H 2 S M i x t u r e (g2 c m - 4 s e c - 1 )

Temp.

Co-15Nb-5A]

Co-15Nb-10A1

600~

2.39 x 10 9

2.10 • 10

700~

4.45 x 1 0 - 9

4.64 x 1 0 - 1 ~

lO

Co-15Nb-15A1

Co-15Nb-15Cr

Gas

Co-15A1

1.79 x 10 -11

8.85 • 10 - J ~

--

6.44 x 10 -11

2.10 • 10 - 8

1.91 x 10-22

800~

3.38 x 10 - 8

3.44 x 10 - 9

1.17 x 10 - 9

3.00 x 10 - 8

2.03 x 10 -12

900~

4.03 x 10 - 7

9.84 • 10 - 8

5.73 x 10 - 7

1.76 x 10 - 7

1.84 x 10 - 7

High-Temperature Corrosion of Co-15a/oNb Alloys

20

121

20 Co-15Nb-5AI

r,I

g

1

Co-15Nb.10AI

900"C

800"C

900*~j

700"C

.~ I0 8000C "g 5

700"C

~

50

100 t i m e ,a see

150

0

200

.

.

.

.

i

50

100

'~

t i m e 'a

i

t

150

,

200

see 1/2

15 900"c

Co- 15Nb. 15AI

g e~

g5

~ 7 8 0 0 ~

{

ooocI I00

200

300

400

t i m e ,~2 s e e ' a

Fig. 2. Parabolic plots of the corrosion kinetics of Co 15Nb-A1 alloys.

the activation energies for C o - 1 5 N b - A 1 were determined from kinetics over the temperature range of 600 to 800~ Activation energies for the corrosion of C o - 1 5 N b - 5 A 1 and C o - 1 5 N b - 1 0 A 1 were the same, with an average value of 24.5 ___1.0 kcal/mole, while that of C o - 1 5 N b - 1 5 A 1 was 37.1 kcal/mole. This indicates that a similar mechanism may exist for the corrosion of low-A1 alloys, but a different mechanism exists for that of high-A1 alloys. Figure 4 shows another Arrhenius plot featuring a comparison of the parabolic rate constants of C o - 1 5 N b - 1 5 A 1 , C o - 1 5 N b - 1 5 C r , and C o 15A1 alloys at various temperatures. Chromium is a well-known oxidation-

122

Shing and Douglass

T 1000 I

1 0 -s

7

1 0

-6

10

-7

10

-8

~

900

800

700

600

I

I

I

I

4-

r..)

m

r b~ 1 0

-9

1:~10 -lo_

1 0 --11

Co-15Nb-15A~ 0

-12

7.0

I

I

8.0

i

1/T

I

9.0

'

(xl0')

I

10.0

i

~

[

1 1.0

i

12,0

Fig. 3. Arrhenius plot of parabolic rate constants for C o - 1 5 N b and C o - 1 5 N b - A 1 alloys.

resistant addition to most commercial alloys. The reason for including C o - 1 5 N b - 1 5 C r was to see whether any beneficial effect on the corrosion resistance occurs with the addition of Cr and to compare this effect with that obtained from the addition of A1. Moreover, a Co-15A1 alloy was studied in order to see if the high-A1 content was capable of forming enough A1203, which provides excellent corrosion resistance for the binaries. Obviously, aluminum is more efficient in reducing the corrosion rates than chromium; Co-15A1 is very effective in reducing the corrosion

123

High-Temperature Corrosion of Co-15a/oNb Alloys

T 1000

900

I

1 0 "~

~

800

I

700

I

600

I

I

1 0 -o

t

%

1 0 -~'

I

~~

" ~

Co-15Nb-t5Cr

m-lO*

T

11%1%%

0

1 0 -*'

I:~10 -,o.

I I I i

10 -11.

10

-12

7.0

'

I

8.0

I/T

i

I

9.0

Co-15A1 '

i

10.0

I

I

1 1.0

(xlO 4) OK-'

I

12.0

Fig. 4. Arrhemus plot for the corrosion kinetics of Co-15Nb-15A1, Co-15Nb-15Cr, and

Co- 15A1.

rates over the temperature range of 700 800~ but has no better effect at 900~ The influence of aluminum concentration on the corrosion rate of C o - 1 5 N b - A I alloys is shown in Fig. 5. The corrosion rates decreased with increasing Al content at all the temperatures except 900~ The aluminum content had virtually no effect on the corrosion of alloys studied at 900~ The corrosion rate of C o - 1 5 N b - A 1 at 600~ was virtually the same as

124

Shing and Douglass

10 900~

10

I0

T

800~

in 10

T

700~

O

%1o 600~ 10 -,1

10 -|r

0

I

I

4

I

I

8

I

I

12

I

16

Atomic percentage of A1 Fig. 5. Effect of A1 additions on the corrosion kinetics of Co-15Nb-A1.

that of C o - 1 5 M o - A 1 ternaries in mixed gases. This indicates that the addition of A1 provided similar benefits in corrosion resistance in both binaries. But A1 improves the corrosion resistance of C o - 1 5 N b - A I alloys more effectively than that of C o - 1 5 M o - A 1 alloys at 700~ This effect is shown in Fig. 6. However, the addition of A1 to C o - 1 5 N b alloys does not provide for better corrosion resistance than that obtained for C o - 1 5 M o A1 at 800~ In fact, it provided less corrosion resistance than that of C o - 1 5 M o - A I ternaries at 900~

Scale Morphologies and Phase Constitution The nature of the scales formed on all the alloys can be divided into two groups corresponding to two different temperatures ranges: (1) corro-

125

High-Temperature Corrosion of Co-15a/oNb Alloys

10 "~

I

700oc

I 0 -e

T9 r

10-'

7 E

xAI

e,io

Q. v

C o - 15Nb-xAl

10 "J~I

10 -It 0

'

4' ' 8~ ' 1:2 ' 16/ ' 201 , 24 Atomic percentage of AI

Fig. 6. Effect of Mo and Nb on the corrosion rate of Co-Al-base ternaries,

sion from 600 to 800~ separately.

and (2) corrosion at 900~

Each will be discussed

Corrosion at 600-800~

A duplex-scale structure formed on all C o - 1 5 N b - A I alloys at 600800~ Cross sections of the corroded alloys are shown in Figs. 7-10. The outer layer consisted of Co9S8 and/or CoSl+x. The inner layer was complex and heterophasic, containing mainly a double sulfide, CoNb3S6, mixed with Nb 3$4 and some partially corroded intermetallic phase, NbCo3. In fact, the X-ray diffraction pattern of CoNb3 $6 is nearly identical to that of CoNb2S4, so the identification of both sulfides is extremely difficult. Thus, throughout this paper, the formula CoNb3S6 always means CoNb3S6, CoNb2S4, or both. An unknown phase was found by XRD in the inner layer for all alloys. It is very possible that the unknown phase is

126

Shing and Douglass

@

,'i~, ,,~

Fig, 7. Cross section of Co-15Nb-15A1 corroded 30 hr at 600~

Fig. 8. BE image of the scale formed on Co-15Nb-5A1 corroded 4 hr at 800~

High-Temperature Corrosion of Co-15a]oNb Alloys

~,i.'

127

'~ = m ' q ~ = ' ! ~ ' T . -"

l .,re. n

L

Id

"

.. g,..

"~

Fig. 9. BE image of the scale formed on Co-15Nb 10A1corroded at 800~ for 6 hr.

128

Fig. 10. BE image of the scale formed on C o - 1 5 N b

Shing and Douglass

15AI corroded at 800~ for 6 hr.

High-Temperature Corrosion of Co-15a/oNb Alloys

129

a ternary sulfide, A105NbS2,13 but, unfortunately, there is no X-ray diffraction pattern available. The odor of HzS was detected from the samples during the X-ray analysis of the scales, suggesting the additional presence of A12S3 in the scales. XRD results showed little AI203 in the inner scale of Co-15Nb-15A1, but no A1203 was detected in the scales of alloys containing up to 10 at.% A1. This is probably because the amount of A1203 was too small to be detected by XRD. Inspection of the inner layer of scales revealed that the a-Co and the CoAl phases in the base alloy are attacked preferentially, and the intermetallic phase, NbCo3, corroded at a considerably slower rate than that of the other two phases. The localized attack on the a-Co and CoAl phases was more pronounced as the A1 content increased and the temperature decreased. A typical localized attack is shown in Fig. 7, which is the scale image of Co-15Nb-15AI corroded at 600~ for 30 hr. Furthermore, the better corrosion-resistant phase, NbCo3, was corroded partially in the inner layer. Figure 8 shows the BE image of the scale formed on C o 15Nb-5A1 reacted at 800~ for 4 hr, and the bright phase is the uncorroded NbCo3 retained in the inner scale. A high-magnification BE image of Co-15Nb-10A1 corroded at 800~ for 6 hr is shown in Fig. 9. The dark area is complicated and contained Co, Nb, S, and a large amount of A1. The bright phase, near the interface between the outer and inner layers, is uncorroded NbCo3, while the gray area surrounding it is corroded NbCo3, which contained mainly Nb and S with a very small amount of Co and A1. This gray region is believed to be the Nb3 $4 phase with Co and A1 intercalated into it. The other gray regions containing a large amount of Co are believed to be the double sulfide CoNb3S 6. It is interesting to find that the boundary between the dark region and the gray region was needlelike. It is not possible to identify the needles since their sizes were very small. Unlike the needles, many fine whiskers, surprisingly, were found in the black region of the inner layer on Co-15Nb-15A1 reacted at 800~ for 9 hr. This morphology can be seen clearly in a high-magnification BE image seen in Fig. 10. In fact, the whiskers were found also on the other samples; however, their presence was not as clearly delineated as that shown in Fig. 10. These whiskers were found to be enriched with A1 and S and have some amounts of Co and Nb. Due to the extremely fine shape of the whisker, its identification was not possible. Corrosion at 900~

All the alloys studied formed a very thick and porous scale at 900~ A typical image is shown in Fig. 11. X-ray diffraction results showed that the outer region of the scale contained CoNb3S 6 and C09S8, and the inner

130

Shing and Douglass

9

.'r

?

~i~ ,~ III

,': ,IdO'ilm

'

Fig. 11. The scale morphology of Co-15Nb-15A1 reacted 0.5 hr at 900~

L,d

v

C

5"

fg

Fig. 12. Hlgh-magmfication BE image of the scale formed on Co 15Nb-15A1 reacted 0.5 hr at 900~

High-Temperature Corrosion of Co-15a/oNb Alloys

131

region consisted mainly of CoNb 3S6 and, surprisingly, a-Co containing 90% of Co. A very small amount of Nb3S 4 was also found in the inner portion of the scale. Figure 12 shows a high-magnification BE image of the scale formed on Co-15Nb-15A1 at 900~ The dark area represents pores, the bright phase is Co-rich metal, and the gray phase is double sulfide. Marker Studies Inert-marker experiments were carried out to study the corrosion mechanism of the C o - N b - A 1 alloys. The Pt marker was found to be located at the interface between the outer layer and the inner layer. Figure 13 shows the marker position clearly in the scale formed on C o - 1 5 N b 10A1 after corrosion at 800~ for 52 hr.

i

Fig. 13. C o - 1 5 N b

"

10A1 corroded 52 hr at 800~

showing the marker location.

132

Shing and Douglass

DISCUSSION Several features of the high-temperature corrosion of Co-15Nb-A1 alloys were observed as reported above. These features can be summarized as follows: (1) the corrosion resistance increased with increasing Al content; (2) A1203 was detected only in the scale on Co-15Nb-15A1 alloys; (3) niobium oxide was absent in the scale; (4) the formation of Nb3S4 was detected in the inner layer; (5) there was an unknown phase within the scale; (6) completely different scale morphology, the presence of e-Co in the scale, and virtually no Al-benefit effect on the corrosion of alloys at 900~ (7) the presence of c~-Co in the scale product formed at 900~ and (8) completely different scale morphology at 900~ compared to that of lower temperatures. Two features deserve mention that appear to be diametrically opposite. First, corrosion rates of Co-15Nb-A1 alloys decreased with increasing A1 content, and second, the NbCo3 phase in the alloys was found to corrode very slowly. Thus, it is tempting to say that decreasing corrosion rates are attributed to increasing amounts of the corrosion-resistant phase, NbCo3. However, as shown in Table III, the volume fraction of NbCo3 decreased with increasing A1 content in the alloys; thus, the inherently low corrosion rate of NbCo 3 probably has little or nothing to do with the improved corrosion rate observed. The most likely explanation of the reduction in corrosion rate in high-A1 alloys is that other reaction products form in the scales, and these are more protective than those formed in low-A1 alloys. A small amount of A1203 was detected in the scales formed on the Co-15Nb-15A1 alloy, and it never formed a continuous layer. Similarly, AI203 was found only on Co-15Mo-A1 alloys containing > 15 at.% AI corroded at high temperatures with the same gas mixture used with Co-15Nb-A1 alloys. The formation of A1203 on Co-15Mo-15A1 was attributed to the amount and the distribution of the third phase (CoAl) in those alloys; however, the slow growth rate of A1203 still enables the fast-growing cobalt sulfide to easily outgrow it. 11 Thus, the same reasons may contribute to the absence of A1203 on Co-15Nb-5A1 and C o - 1 5 N b 10A1 because of the very small amount of CoAl in these two alloys, 0% and 13%, respectively. However, it was not possible to form an A1203 layer on C o - 15Nb- 15A1 because the amount of CoAl in the alloy was not large (27%). Moreover, many tiny whiskers were found in the dark region of the inner layer shown in Fig. 10. Although the size of the whiskers was too small to analyze accurately, EDX results did show that there was a large amount of A1 and a low concentration of other elements in this region. Whisker-shaped A1203 was also found on the scales of Fe-30wt.%

High-Temperature Corrosion of Co-15a/oNb Alloys

133

Nb-7 wt.% A1 and - l 0 wt.% A1 alloys at 7 0 0 - 8 0 0 ~ 14 However, those Al~O3 whiskers were found on the top of the outer scale. The reason for the A1203 whisker formed in the inner layer is unknown. Apparently, the A1203 whiskers are nonprotective. In a previous study ~ the improvement in the corrosion resistance of the Co-15Mo-A1 alloys with increasing A1 content was attributed to the presence of the spinel phase, Alo55M02S4, and A1203 in the inner layer. The corrosion resistance of Co-15Nb alloys was improved by increasing additions of A1. No double sulfide of the A1-Nb-S system comparable to the spinel phase, AI0.55Mo~S4, of the A1-Mo-S system could be identified by XRD because there is no,X-ray diffraction pattern available. But there was an unknown phase present in the scales formed on Co-15Nb-A1 ternaries regardless of alloy composition and temperature. It would be expected that a similar compound exists in the A1-Nb-S system because of the same layered structure characteristic of both NbS 2 and MoS2. A ternary compound, Alo.sNbS~, was reported by Matukhin e t al. 13 This compound is formed by the intercalation of A1 into layered NbS2, and it retains the hexagonal structure of NbSz. Thus, it is very possible that the unknown phase in this study is Alo.sNbS2. The preferential intercalation of aluminum in the inner scale will block the transport of cobalt through this region and will decrease the growth rate of the outer scale. As well, A10.sNbS~ may reduce the corrosion by reducing the cross-sectional area for diffusion since it did not form a continuous layer. It is surprising that NbS2, which was observed in the corrosion of C o - N b binary alloys,6 was not found in the scale formed on Co-15Nb-A1 alloys presently studied. Instead, Nb3 $4 was the only sulfide phase present except for double sulfide(s) formed in the inner layer of Co-15Nb-A1 ternaries. The formation of Nb3 $4 was observed also in the corrosion of Fe-Nb-A1 alloys. ~2,~5 The appearance of Nb3S 4 might be due to the presence of A1 in the alloys, which destabilized the layered NbS2 structure to form Nb3S4 and/or Alo.sNbS2. In fact, the formation of Nb3S 4 is unpredictable due to the fact that NbS 2 is the only sulfide, for which thermodynamic data are available. In addition, Nb3S 4 is apparently more stable than NbS2: its dissociation sulfur partial pressure is lower than that of NbS2. Thus, it is very likely that the sulfur was blocked efficiently in the outer region of the inner layer by the formation of A12S3, and its activity decreased to a value to where Nb3S 4 is stable. Some Nb3S4, shown in the gray region of Fig. 9, located next to the unattacked NbC%, indicates this possibility. It is worth noting also that NbO 2 was present in corrosion scales of C o - N b binaries but iwas absent in this study. The same arguments mentioned above can be applied to explain the reason for the absence of

134

Shing and Douglass

NbO2. In this case, oxygen is effectively blocked by the formation of A1203. Thermodynamically, A1203 is much more stable than NbO2. For example, the standard free energy of formation of A1203 is -217 kcal/mole 02 and that of NbO2 is -146 kcal/mole O2 at 1000 K. 14 Therefore, the oxygen activity should be too low to form NbO2 after the formation of A1203. However, the amount of A1203 formed during corrosion of C o 15Nb containing up to 10 at.% A1 is believed to be very small due to its absence in X-ray diffraction analysis. Consequently, it would be more reasonable that both AlzS 3 and the unknown phase in the inner layer also contribute to the blockage of oxygen. The absence of NbO2 in the scale on Co-15Nb-A1 alloys may also be attributed to the effect of A1 on the activity of Nb in the alloys. Since the thermodynamic properties, such as activities, of the Co-Nb-A1 ternary system are not available, ternary values can be determined by using binary values. A power series is used most often to express the thermodynamic properties of alloys, i.e., the Margules equation. The excess free energy for a binary system can be described generally by a third-order Margules equation: 16 Gel2 = x l x 2 ( A 2 1 x 1 q.- A12x2)

(1)

The values of G ~ (in cal/g- atom) for Co-Nb, ~7 Nb-A1, is and Co-A117 binary alloys are as follows: G e(Nb_ Co) = - 1700XNbXco

(2)

Ge(Nb-AI) = --XNbXA,(51000- 16.2T)

(3)

Ge(Co-A1) = XNbXCo[( -- 40000 + 63.783T - 0.08625T 2 + 0.31743 x 10-4T3)xco - 25000Xal]

(4)

where T is the temperature. Thus, the excess partial molar free energy of Nb((7~b) in Co-15Nb-A1 alloys can be obtained by using the binary coefficients in Eqs. (1)-(4). 16 Taking a temperature of 800~ for example, the result for G~b is a~'qb

=

--

1700x~o -- 33617.4x~1 -- XCoXAl(3660.9 + 13313XA1)

(5)

In ~Nb

(6)

Recalling that --e

--

G N b -- R T

where ~Nb is the activity coefficient of Nb. Therefore, the value of ])Nb at Xco=0.75, XAI=0.1 (Co-15Nb-10A1, for example) is 0.456, which is lower than that of Co-15Nb without A1, 0.562. The value of 7Nb decreases with decreasing temperature and with increasing A1 content in C o - l 5Nb-

High-Temperature Corrosion of Co-15a/oNb Alloys

135

A1 alloys. The decrease of 7Nb means that A1 additions lower the activity of Nb in alloys. Thus, the decreasing activity of Nb by AI might also contribute to the inhibition of NbO 2 formation. The reduction in corrosion rate of alloys presently studied is believed to be due to the unknown phase, which is considered to be A1o 5NbS2. The intercalated A1 ions in this phase would effectively block the outward diffusion of cobalt ions in the inner layer. The formation of Nb3 $4 might play an important role in reducing the corrosion rate also. Therefore, it is of interest to examine the crystal structure of Nb3 $4. Nb3 $4 has a hexagonal crystal structure. In contrast to NbS2, Nb3 $4 forms a three-dimensional network structure of edges and faces sharing NbS 6 octahedra instead of forming a layered structure. The niobium atoms form zigzag N b - N b - N b chains running along the c direction. 19 The interchain distance between the Nb atoms is short, indicating metallic bonding, while the distance between Nb atoms in neighboring chains is considerably larger, suggesting quasione-dimensional behavior. Thus, a projected view of Nb3S 4 along the hexagonal e-axis can be shown in Fig. 14. An interesting feature of the Nb3S 4 structure is the presence of large infinite channels parallel to the c-axis. These channels appear to be empty. The space of these empty channels is wide enough for ternary element insertion in the lattice. Huan and Greenblatt 2~ have reported a large number of ternary niobium sulfides AxNb6S8 with A = Bi, In, Zn, Pb, T1, etc., which were formed by ternaryelement insertion into the Nb3S 4 lattice. Therefore, it is believed that A1 and Co ions can insert easily into the empty channels in the Nb3 $4 lattice. It is worth noting also that the low filling of the channels is due to lattice defects. 21 The x values of the AxNb6S 8 studied by Huan and Greenblatt 2~ are close to 2.0 for monovalent ions, approach ~ 1.0 for bivalent, and ~0.67 for trivalent ions. Thus, a smaller concentration of aluminum ions (3 +) than cobalt ions (2+) can be filled into Nb3S 4. The intercalated ions occupying the hexagonal channels may further block the transport of cobalt ions through the inner layer and reduce the corrosion rate of alloys. Regarding the diffusion mechanism, the marker studies showed that the Pt marker was located at the inner/outer layer boundary. This clearly indicates that the growth of the outer cobalt sulfide is due to outward diffusion of cobalt ions, while the inner layer grows by means of inward diffusion of sulfur. However, the latter conclusion is not completely correct. 22 This is because of the multiphase nature of the alloy structure, i.e., the intermetallic compound in this region of the scale might prevent the marker from making direct contact with the alloy surface. Similarly, the presence of multiphase corrosion products in the inner layer complicated the case also. Therefore, the mechanism of transport within the inner layer is complex and cannot be conclusively determined by the present marker experiments.

136

Shing and Douglass

o o Nb oQS Fig. 14. Schematic diagram of the structure of Nb3S4, projection along the hexagonal c-axis.

C o - 1 5 N b - A 1 alloys containing up to 15 at.% A1 corroded at 900~ formed a single layer which consisted of mainly CoNb3S 6 and a-Co. The scale was porous. A similar behavior was observed in the scale formed on C o - M o binaries after corrosion at 900~ 9 In that case, two phases, one of which is Co-rich C o M o 2 S 4 and the other of which is almost pure metallic Co, formed a layer of lamellar structure. The lamellar structure of C o M o 2 S 4 and metallic Co is attributed to the eutectoid decomposition of an unknown double C o - M o sulfide. However, the lamellar structure is absent in the present study. The reason for the presence of metallic cobalt is not clear. Nevertheless, a eutectic point exists for a liquid solution of C o - S at 880~ at which the C o - S liquid solution kvill decompose into

High-Temperature Corrosion of Co-15a[oNb Alloys

137

e-Co and cobalt sulfide. It is possible that niobium sulfide has a large solubility in cobalt sulfide instead of in e-Co. Thus, niobium sulfide reacts with cobalt sulfide to form the double sulfide CoNb3S6, and e-Co remained unreacted. This possibility is supported by the fact that a much higher corrosion rate, which sould be observed while a liquid phase is present in the scale of 900~ is in agreement with the abnormally higher corrosion rate obtained. Besides, the small amount of Co metal, which was actually observed (bright phase) in Fig. 12, was consistent also with that produced by eutectic decomposition. The other possible reason concerns the phase equilibrium established between a-Co and CoNb3S6. However, it is difficult to analyze the phase relationships due to the lack of a ternary phase diagram of the C o - N b - S system. Thus, it is reasonable to assume that aluminum has no effect on the corrosion resistance of Co-15Nb-A1 alloys since a possible liquid phase formed on the alloys at 900~ CONCLUSIONS 1. The parabolic rate law was followed for the corrosion of C o - 15Nb alloys containing up to 15 at.% A1 in a H ~ - H 2 0 - H 2 S gas mixture at 600-900~ A duplex scale formed on all alloys regardless of composition and temperature. The outer layer was mainly cobalt sulfide(s). The inner layer was heterophasic containing a cobalt-niobium sulfide, CoMo3S6, mixed with Nb3S4, A12S3, and some partially corroded particles of the intermetallic phase, NbCo3. 2. A1203 was found only on Co-15Nb-15A1. The absence of A1203 in the low-A1 alloys is attributed to the amount and the distribution of the third phase (CoAl) in these alloys and the slow growth rate of A1203 which enables the fast-growing cobalt sulfide to easily outgrow it. 3. NbS2 and NbO2, which were present in the scale of C o - N b binaries were absent on Co-15Nb-A1 alloys. This is because A1 reduced the activity of Nb in alloys and formed also extremely stable A1203 and AI~S3 phases that efficiently blocked the inward diffusion of oxygen and sulfur, respectively. 4. The formation of an unknown phase (possibly A10.sNbS2) and Nb 3S4 contribute to the significant reduction of the corrosion rate. This is because of the intercalation of A1 into the lattices to further block the transport of cobalt ions. 5. The corrosion rates of Co-15Nb-A1 alloys at 900~ were abnormally high. The scales formed consisted of CoNb3S6 and almost pure metallic Co. A possible eutectic decomposition of a liquid solution of Co-S may be responsible for this phenomenon and for the abnormally high corrosion rate also.

138

Shing and Douglass

ACKNOWLEDGMENTS This work was supported by a grant from the Electric Power Research Institute, Palo Alto, California. The interest and support of Dr. John Stringer, technical advisor, is gratefully appreciated. REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22.

K. Natesan, Corrosion 41, 646 (1985). S. Mrowec and K. Przybylski, High Temp. Mater. Proc., 6, 1 (1984). H. S. Hsu, Oxid. Met. 28, 213 (1987). K. N. Stratford and P. K. Datta, Mater. Sci. Tech. 5, 765 (1989). W. Kai, D. L. Douglass, and F. Gesmundo, Oxid. Met. 37, 189 (1992). C. C. Shing, D. L. Douglass, and F. Gesmundo, Oxid. Met. 37, 167 (1992). Y. R. He, D. L. Douglass, and F. Gesmundo, Oxtd. Met. 37, 217 (1992). W. Kai, D. L. Douglass, and F. Gesmundo, Oxid. Met. 37, 389 (1992). C. C. Shing, D. L. Douglass, and F. Gesmundo, Oxid. Met. 37, 441 (1992). Y. R. He, D. L. Douglass, and F. Gesmundo, Oxid. Met. 37, 413 (1992). C. C. Shing and D. L. Douglass, Oxid. Met. 40, 155 (1993). G. Wang, D. L. Douglass, and F. Gesmundo, Oxid. Met. 35, 279 (1991). V. L. Matukhin, I. A. Satin, V. F. Shamrai, and G. M. Leitus, Soy. Phys. Solid State 27, 117 (1985). JANAF Thermodynamical Tables, 3rd ed., 1985 supplement, from J. Phys. Chem. Ref. Data 14 (1985). W. Kai and D. L. Douglass, Oxid. Met. Oxid. Met. 39, 415 (1993). N. A. Gokcen, Statistical Thermodynamics of Alloys (Plenum Press, New York, 1986). L. Kaufman and H. Nesor, Met. Trans. 6A, 2115 (1975). L. Kaufman and H. Nesor, Calphad 2, 325 (1978). A. F. J. Ruysink, F. Kadijk, A. J. Wagner, and F. Jellinek, Aeta. Crystallogr. Sect. B 24, 1614 (1968). G. Huan and M. Greenblatt, Mater. Res. Bull. 22, 943 (1987). R. Sch611horn and W. Sehramm. Z. Naturforsch. Teil B 34, 697 (1979). B. Gleeson, D. L. Douglass, and F. Gesmundo, Oxid. Met. 31, 209 (1989).