International Journal of Hydrogen Energy 32 (2007) 3778 – 3788 www.elsevier.com/locate/ijhydene

Growth of Cr-Nitrides on commercial Ni–Cr and Fe–Cr base alloys to protect PEMFC bipolar plates M.P. Brady a,∗ , H. Wang b , B. Yang a , J.A. Turner b , M. Bordignon c , R. Molins c , M. Abd Elhamid d , L. Lipp e , L.R. Walker a a Oak Ridge National Laboratory, Oak Ridge, TN 37831-6115, USA b National Renewable Energy Laboratory, Golden, CO 80401, USA c Ecole des Mines de Paris, Centre des Matériaux, UMR CNRS 7633 B.P.87 F91003 EVRY, France d General Motors Technical Center, Warren, MI 48090, USA e FuelCell Energy, Inc., Danbury, CT 06813, USA

Available online 10 October 2006

Abstract Nitridation of Cr-bearing alloys can yield low interfacial contact resistance (ICR), electrically conductive and corrosion-resistant CrN or Cr2 N base surfaces of interest for a range of electrochemical devices, including fuel cells, batteries, and sensors. This paper presents results of exploratory studies of the nitridation of commercially available, high Cr (30–35 wt%) Ni–Cr alloys and a ferritic high Cr (29 wt%) stainless steel for proton exchange membrane fuel cell (PEMFC) bipolar plates. A high degree of corrosion resistance in sulfuric acid solutions designed to simulate bipolar plate conditions and low ICR values were achieved. Oxygen impurities in the nitriding environment were observed to play a significant role in the nitrided surface structures that formed, with detrimental effects for the Ni–Cr base alloys, but beneficial effects for the stainless steel alloy. Positive results from single-cell fuel cell testing are also presented. 䉷 2006 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved. Keywords: Nitride; Bipolar plate; Fuel cell; Oxidation; Nitridation

1. Introduction Proton exchange membrane fuel cells (PEMFCs) are of interest for power generation due to their high efficiency and near-zero emissions [1–5]. A key component for PEMFCs is the bipolar plate, which serves to electrically connect the anode of one cell to the cathode of another in a stack to achieve a useful voltage. It also separates and distributes reactant and product streams. To accomplish this, flow-field grooves are manufactured into the faces of the plates. Graphite is the benchmark material for bipolar plates because of its electrically conductivity and corrosion resistance in the aggressive anode and cathode PEMFC environments (typically 60–80 ◦ C acidic conditions) [1–5]. However, the brittleness and relatively high gas permeability of graphite necessitates the use of thick plates (> 2.5 mm), which lowers the volumetric power ∗ Corresponding author. Tel.: +1 865 574 5153; fax: +1 865 241 0215.

E-mail address: [email protected] (M.P. Brady).

density of the fuel cell stack. Machining of flow fields into graphite plates is also expensive, making graphite impractical for most wide-scale commercial uses. Developmental bipolar plate materials under investigation include graphite/carbonbased composites [6–9], polymer-based composites with conductive graphite/carbon fillers [10–15], and metallic alloys with/without surface treatments or coatings [16–30]. Metallic alloys such as stainless steels are of interest for bipolar plates [e.g. 5,18,19] because they are amenable to lowcost/high-volume manufacturing methods such as stamping, offer high thermal and electrical conductivities, have low gas permeability, excellent mechanical properties, and can be readily made in foil form ( 0.1 mm thick) to achieve high volumetric power densities. From a cost perspective, the rising cost of Ni and Mo has rendered the higher Ni/Mo-containing austenitic stainless steels and Ni(Mo)-base alloys too expensive for some PEMFC applications, particularly transportation [31]. In this regard, ferritic stainless steels (based on Fe–Cr) are of particular interest as a bipolar plate material, although

0360-3199/$ - see front matter 䉷 2006 International Association for Hydrogen Energy. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.ijhydene.2006.08.044

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

the use of more expensive alloys as claddings or coatings on inexpensive substrates is currently being explored for bipolar plate applications [32,33]. From a performance perspective, the drawbacks of stainless steels and related Cr-bearing alloys are borderline corrosion resistance in the PEMFC environment and high interfacial contact resistance (ICR) from the oxides typically formed on their surface [e.g. 27–29]. To overcome these limitations, corrosionresistant coatings or surface treatments with low ICR/high electric conductivity are needed. Conventional coating deposition methods, however, have thus far not proven sufficiently viable due to pin-hole defects [34], which can result in local corrosion and metallic ion contamination of the membrane. Recently, it was demonstrated that thermally grown Cr nitrides (CrN/Cr2 N) formed by nitridation of a model Cr-bearing alloy, Ni–50Cr weight percent (wt%), showed good behavior in PEMFC bipolar plate environments [31,35–38]. The nitrided Ni–50Cr exhibited low anodic current densities, stable and very low interfacial contact resistance, and little metallic ion dissolution under simulated PEMFC anodic and cathodic operating conditions. Single-cell fuel cell testing under static [36] and drive-cycle conditions [31] also indicated good behavior, with no increase in ICR and virtually no metallic ion contamination of the polymer membrane. However, this alloy is of limited ductility and too expensive for many PEMFC applications. The goal of the present work was to explore the nitridation of several commercially available Ni–Cr and Fe–Cr base alloys, and evaluate their electrical properties and corrosion resistance relative to use as PEMFC bipolar plates. 2. Experimental 2.1. Materials and characterization Two commercial high-Cr, Ni–Cr base alloys and a high-Cr ferritic stainless steel were selected for study: HASTELLOY䉸 G-30䉸 , nominal composition of Ni–30Cr–15Fe–5.5Mo–2.5W– 5Co–2Cu–1.5Nb–1.5Mn–1Si–0.03C; HASTELLOY䉸 G-35姠, nominal composition of Ni–33.2Cr–8.1Mo–2Fe–0.6Si–0.3Cu– 0.05C, and AL 29-4C䉸 , nominal composition of Fe–29Cr– 4Mo–0.3Ni–0.5Mn–0.03P–0.01S–0.4Si–0.03Co–0.02C–0.02N –0.5(Ti+Nb), all in wt%. The Ni–Cr base alloys were selected based on the work of Rubly and Douglass [39], which indicated that the transition from internal to external Cr-nitride formation in binary Ni–Cr alloys occurred in the range of 30–35 wt% Cr. The ferritic alloy with high Cr content was selected based on earlier studies [40] which showed beneficial effects of nitridation on type 446 MOD-1 stainless steel, a highCr ferritic alloy with Ti microalloying additions. (Chromium levels above 29–30 wt% are not viable in Fe–Cr base alloys due to the increased potential for  phase formation, which causes embrittlement and possibly other detrimental effects in a bipolar plate application). Disk or rectangular-shaped coupons with dimensions 12 to 24 mm and ∼1 mm thickness were cut by electrical discharge machining and prepped to a 240 grit surface finish using SiC paper. The coupons were nitrided using an

3779

Al2 O3 -tube vacuum furnace. The furnace was first evacuated to ∼10−5 Pa (∼10−7 Torr) vacuum, backfilled with 96%N2 –4%H2 (volume percent) gas (referred to as N2 –4%H2 for conciseness) to 101 kPa (1 Atm), and then heated to the nitridation temperature over 4 to 6 h. Unless otherwise specified, the N2 –4H2 gas was static rather than continuously flowing during the nitridation run to prevent the constant introduction of trace oxygen and water vapor impurities present in the gas, i.e. a closed system was used. The furnace system used was equipped with a pressure-relief valve to withstand the pressure increase during heating associated with the expansion of the N2 –4H2 gas with increasing temperature. Selected samples were also nitrided in this system using flowing N2 –4H2 that was purified with a commercially available molecular sieve cartridge in line with the gas cylinder to remove oxygen and water vapor impurities from the gas. Nitrided microstructures were analyzed by scanning electron microscopy (SEM), electron probe microanalysis (EPMA) using pure element standards for metals, Al2 O3 for oxygen, and BN for nitrogen, scanning/transmission electron microscopy (STEM/TEM), and Auger electron spectroscopy (AES). For transmission electron microscopy, samples were prepared using a tripod and a precision ion polishing system (PIPS). Then they were observed by the STEM technique using a high angle annular dark field (HAADF) detector, which gives images with topographical and chemical contrast. EDS analysis was done using a nanoprobe, and are considered semi-quantitative in part due to the empiric thickness correction applied (100 nm). The AES depth profiles were obtained by sputtering with 3 keV argon ions and a current density of around 50 A/cm2 . During the sputtering, the background partial pressure of Ar in the chamber was typically 1.3× 10−6 Pa (1 × 10−8 Torr). Based on previous measurements, a reasonable estimate of the sputtering rate was ∼50 nm/ min. 2.2. ICR evaluation ICR values were obtained using two pieces of conductive carbon paper that were sandwiched between the sample and two copper plates. A current of 1.000 A was provided via the two copper plates and the total voltage drop was registered. The contact pressure was then systematically increased. In this manner, the total resistance dependency as a function of contact pressure was determined. Corrections were made for the resistance of the carbon paper/copper plate interfaces (RC/Cu ) by calibration, and the obtained values were divided by two to obtain ICR values for a single carbon paper/nitride interface. For samples examined after corrosion exposure, in which only one coupon face was exposed to the environment, the calculated ICR values were not divided by two and thus represent both sets of carbon paper/nitride interfaces. Further details of this technique are provided in Ref. [27]. 2.3. Corrosion evaluation The corrosion behavior was evaluated by polarization in either pH 3 sulfuric acid at 80 ◦ C or 1 M sulfuric acid

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

+2 ppm F—at 70 ◦ C. The solutions were sparged with air to simulate the cathodic environment or with Ar–4H2 (pH 3 sulfuric)/pure H2 (1 M sulfuric +2 ppm F− ) to simulate the anodic environment. All potentials in this paper are presented relative to the standard hydrogen electrode (SHE) unless otherwise noted. Further details of the corrosion test technique are available in Refs. [37,40].

Potential (mV vs SHE)

3780

2.4. Single-cell fuel cell testing

3. Results and discussion 3.1. Nitrided commercial Ni–Cr base alloys Polarization data for nitrided G-30䉸 and G-35姠 coupons in pH 3 sulfuric acid at 80 ◦ C are shown in Fig. 1. Data for

Potential (mV vs SHE)

G-30 Metal G-35 Metal 310 Metal

1.E-07

1.E-06

1.E-05

1.E-04

Current Density (A/cm ) 500 Nit. Ni-50Cr 400 Nit G-35 300 Nit G-30 200 100 0 -100 -200 -300 -400 -500 1.E-09 1.E-08

Goal

1.E-07

310 Metal

1.E-06

1.E-05

1.E-04

2

(b)

(c)

Goal

2

(a)

Potential (mV vs SHE)

Single-cell fuel cell testing was conducted for nitrided anode and cathode plates machined from the G-35姠 alloy. Two tests were conducted. The first utilized ∼10 cm × 10 cm anode and cathode plates with a ∼50 cm2 active area based on a double serpentine flow field groove pattern machined into the surface of the alloy prior to nitridation. The fuel cell testing was performed at 80 ◦ C using a Gore 5510 membrane electrode assembly (MEA), with a Pt loading of 0.4 mg/cm2 , and H2 and air at 25 psig and 2/2 stoichiometry for anode/cathode. A constant current of 0.2 A/cm2 was used. The relative humidity (RH) was cycled from 40% to 100% to exacerbate leaching of F− from the MEA to produce conditions aggressive to the anode and cathode plates. The F− release rate was 9 × 10−7 to 3.4 × 10−6 g/cm2 /h. Post test evaluation included X-ray fluorescence analysis (XRF) of the MEA. The second test utilized ∼10 cm × 10 cm anode and cathode test plates with a ∼25 cm2 active area based on a single serpentine flow field groove pattern machined into the surface of the alloy prior to nitridation. The MEA was placed between the nitrided G35 plates. The backs of the nitrided plates were in contact with blank graphite plates, compressed against Au plated Cu current collectors. The MEA was made using a ∼50 m thick proprietary membrane and 40% Pt/C catalyst with a loading of 0.21 mg/cm2 Pt on the anode and 0.47 mg/cm2 Pt on the cathode. The cell was operated at a constant temperature of 60 ◦ C with fully saturated air and hydrogen. The current density was kept constant at 0.4 A/cm2 with a hydrogen stoichiometry of 1.2 and an air stoichiometry of 2. Nitridation of the machined G-35姠 alloy test plates was conducted in a molybdenum vacuum furnace using flowing N2 –4H2 that was purified with a commercially available molecular sieve cartridge in line with the gas cylinder to remove oxygen and water vapor impurities from the gas. The furnace was evacuated to ∼7 × 10−4 Pa (5 × 10−6 Torr) vacuum, backfilled with N2 –4H2 gas, and then heated to 1100 ◦ C in 1.5 h, held at 1100 ◦ C for 8 h, and then furnace cooled. Two separate runs were used to nitride the anode and cathode test plates, i.e. one run for the 50 cm2 active area plates for the RH cycled test at 80 ◦ C and 0.2 A/cm2 and a second run for the 25 cm2 active area plates for the test at 60 ◦ C and 0.4 A/cm2 .

1000 Nit. Ni-50Cr 900 Nit G-35 800 Nit G-30 700 600 500 400 300 200 100 0 1.E-09 1.E-08

Current Density (A/cm ) 1000 900 800 700 600 500 400 300 200 100 0 1.E-09

8h 6h 2h 24 h

1.E-08

Goal

1.E-07

1.E-06

1.E-05

1.E-04

Current Density (A/cm2)

Fig. 1. Polarization data in pH 3 sulfuric acid at 80 ◦ C: (a) aerated conditions: Nit Ni–50Cr: 1100 ◦ C, 2 h, Static N2 ; Nit G-30/G-35: 1100 ◦ C, 6 h, N2 –4H2 static; (b) Ar–4H2 purged conditions (same nitridation as (a)); (c) aerated conditions, G-35 nitrided at 1100 ◦ C in static N2 –4H2 (Nit. = nitrided).

the nitrided model Ni–50Cr alloy [37], and untreated G-30䉸 , G-35姠, and type 310 stainless steel alloy coupons (nominal Fe–(24–26)Cr–(19–22)Ni–(< 2)Mn–(< 1.5 Si)–(< 0.45 P)–(< 0.3 S)–(< 0.25)C wt%) are also shown for comparative purposes. Anodic current densitites less than ∼1×10−6 A/cm2 up to approximately ∼ + 0.9 V vs SHE under aerated conditions, to simulate the cathodic environment, and from ∼ − 0.2 to +0.1 V vs SHE under H2 -purged conditions, to simulate the anodic environment, are considered sufficiently promising to warrant longer term corrosion evaluation and fuel cell testing (based on experience with nitrided alloy surfaces) [31]. The key considerations are the extent to which the anodic current densities observed in such screening tests correlate with processes detrimental to the MEA, i.e. dissolution of metallic ions that lead to membrane poisoning [e.g. 25], or formation of surface oxidation products that increase ICR [e.g. 27–29].

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788 200 180

ICR (mohm-cm2)

160

G-30 Metal G-35 Metal

140 120 100 80 60

Nit G-35

40

Nit G-30

20 Goal

0 0

(a)

50

100

150

200

Compaction Pressure (N/cm2) 250

200 2x ICR (mohm-cm2)

As such, there is no simple pass/fail level of anodic current densities; values lower than the 1 × 10−6 A/cm2 goal may not be acceptable if they involve dissolution and MEA contamination by the highly detrimental species such as Cr, Fe, and Ni. Values higher than this level may be acceptable if the anodic current densities result from the formation of surface reaction products that do not increase contact resistance (reviewed in Ref. 31). The untreated G-30䉸 , G-35姠, and type 310 stainless steel coupons all exhibited good corrosion resistance under aerated conditions, with the untreated G-30䉸 alloy meeting the anodic current density goal up to +0.85.0.9 V vs SHE (Fig. 1a). This result is consistent with Fan et al. [30], who have identified Ni–Cr base alloys as candidate bipolar plate materials. Nitridation of G-30䉸 and G-35姠 for 6 h at 1100 ◦ C in static N2 –4H2 resulted in a moderate, further improvement in corrosion resistance under aerated conditions, yielding target anodic current densities in the range of the model nitrided Ni–50Cr alloy. Under simulated anodic conditions (Fig. 1b), nitrided G-30䉸 and G-35姠 again exhibited target, low anodic current densities, in the range of those exhibited by nitrided Ni–50Cr, and about one order of magnitude lower (better) than untreated type 310 stainless steel shown for comparison. (Untreated G-30䉸 and G-35姠 were not studied under these conditions). Polarization data under aerated conditions for G-35姠 nitrided over a range of times at 1100 ◦ C in static N2 –4H2 are shown in Fig. 1c. The shapes of the polarization curves were similar, the main difference being the open circuit potential, with higher open circuit potential yielding moderately better behavior. The biggest differences were measured for times of 6 and 8 h of nitridation, respectively, and are considered to be a byproduct of the sensitivity of this alloy (and G-30䉸 ) to small variations in the nitriding environment from run to run, rather than an optimum time/temperature window for nitriding existing between 6 and 8 h at 1100 ◦ C. This issue will be discussed in greater detail in the microstructural characterization subsection of this paper. ICR data for untreated and nitrided G-30䉸 and G-35姠 (1100 ◦ C, 8 h, N2 –4H2 ) are shown in Fig. 2. The untreated G-30䉸 and G-35姠 exhibited ICR values in the range of ∼30.75 m-cm2 at the contact pressures of 100.200 N/cm2 of interest for PEMFC stacks (Fig. 2a). These values are significantly lower (better) than those reported for Fe–Cr-based stainless steel alloys, which typically exhibit ICR values > 100 m-cm2 under these conditions [27–29]. However, they fall short of the target value of ∼10 m-cm2 (reviewed in Ref. [31]). Nitridation of G-30䉸 and G-35姠 resulted in a decrease of ICR to the ∼10 m-cm2 range at a contact pressure of 150 N/cm2 for nitrided G-35姠 and less than 100 N/cm2 for nitrided G-30䉸 . To check for possible increases in ICR after polarization, nitrided G-35姠 was subjected to a 7.5 h hold in aerated 1 M sulfuric acid +2 ppm F− at 70 ◦ C and 0.84 V to simulate aggressive PEMFC cathodic conditions. At this potential, the passive film growth on untreated stainless steels/alloys typically yields a high contact resistance with the carbon paper [e.g. 27–29]. The anodic current density under this condition measured with the nitrided G-35姠 alloy was only ∼0.5 × 10−6 A/cm2 . A small increase in ICR was observed,

3781

150 Polarized As- Nitrided

100

50 Goal 0 0

(b)

50

100

150

200

2

Compaction Pressure (N/cm )

Fig. 2. ICR data: (a) Nit G-30/G-35: 6 and 8 h at 1100 ◦ C in static N2 –4H2 ; (b) G-35 as nitrided (1100 ◦ C, 6 h, static N2 –4H2 ) and after polarization for 7.5 h at 0.84 V vs SHE in Aerated 70 ◦ C 1 M sulfuric acid + 2 ppm F− . Note that in (b) ∼2 × ICR is plotted (Nit. = nitrided).

raising the values for G-35姠 to just above the ∼10 m-cm2 goal at a contact pressure of 200 N/cm2 (Fig. 2b). SEM images of the surfaces of one set of nitrided G-30䉸 and G-35姠 (1100 ◦ C, 8 h, static N2 –4H2 ) coupons are shown in Fig. 3. The surface of nitrided G-35姠 (Fig. 3a) consisted of relatively uniform, micron size grains of a Crnitride phase of composition consistent with the CrN phase, (45–65)Cr–(25–55)N–1Mo–(0.5–3)Ni–(0.5–3)Al atomic percent (at%). Of the 10 points examined by EPMA, only two showed evidence of oxygen (< 5 at%). In contrast, the surface formed on nitrided G-30䉸 (Fig. 3b) was more complex, and consisted of a matrix phase (dark) of composition (50–70) Cr–(10–25) N–(5–20) O–(3–5) Mn–(1–4)Ni–(0–1)[Al, Co, Fe, Nb, Si, W] at%, and a light phase of composition (30–35)Cr–(4–15)N–(30–45)O–(4–6) Mn–(10–15)Ni–(2–4) Fe–(3–4)Mo–(0–3)Al–(0–1)[Co, Cu, Nb, Si, W] at%. It should be noted that EPMA analysis of the as-nitrided surfaces should be considered semi-quantitative due to effects of the surface roughness and the possibility of overlap with the underlying alloy.

3782

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

Fig. 3. SEM secondary electron surface images after 8 h nitridation at 1100 ◦ C in static N2 –4H2 : (a) G-35; (b) G-30.

Mass uptakes for nitrided G-30䉸 and G-35姠 on nitridation at 1100 ◦ C were similar, and on the order of 1.5 mg/cm2 for nitriding times in the 6–8 h range. For G-35姠, 2 h nitridation yielded an uptake of 0.7 mg/cm2 and a 24 h uptake of 2.5 mg/cm2 . Cross-section analysis of both alloys after nitridation indicated extensive internal nitridation in addition to external nitride layer formation. In general, the external Cr-nitride layers were borderline for continuity, best described as semicontinuous, and typically only one grain thick. This is in contrast to the microstructure obtained for nitrided Ni–50Cr, which was fully continuous at the surface and consisted of CrN overlying Cr2 N [36], although extensive internal nitridation was also observed in those alloys as well. Work by Rubly and Douglass [39] showed that the transition from internal to external nitridation in binary Ni–Cr occurred around 30–35 Cr wt%, so the borderline continuity of the external nitride formed on nitrided G-30䉸 and G-35姠 is not surprising, given nominal Cr levels of 30 and 33 wt% Cr, respectively. Nitridation of G-30䉸 and G-35姠 was found to be sensitive to oxygen impurity variations in the nitriding furnace from run

Fig. 4. STEM cross-section images after nitridation for 8 h at 1100 ◦ C in purified, flowing N2 –4H2 : (a) G-30; (b) G-35. The composition data should be considered semi-quantitative.

to run. This is partially a consequence of the high thermodynamic stability of Cr2 O3 , compared to the relative low thermodynamic stability of CrN and Cr2 N, which can make it difficult to obtain solely nitriding conditions. At 1100 ◦ C in N2 –4H2 , oxygen levels of greater than only ∼10.15 ppm are sufficient to oxidize rather than nitride Cr. At lower temperatures, Cr-oxide is further thermodynamically favored over Cr-nitride (the samples were heated in the nitriding gas to 1100 ◦ C over 6 h, and it is likely that oxygen incorporation into the surface occurred primarily during heat up). Sources of oxygen contamination include oxygen and water vapor impurities in the N2 –4H2 gas cylinder, and moisture which may have adsorbed on the inner surfaces of the nitriding furnace when the samples were loaded. The ranges of possible nitrided microstructures encountered from run to run are illustrated in Fig. 4. The nitrided G-30䉸 surface shown in Fig. 4a generally consisted of an outer layer of

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

1.0

400 0.8 300 200

0.6

Cell Voltage (V)

Current Density (mA/cm2)

500

100 0.4 200

(a)

400 600 Time (h)

800

1000

1000

900

800

open circuit voltage

800 700

600

cell voltage

600

clamping adjustment

500

200

400

cell resistance

300 0 (b)

400

400

800

1200

1600

2000

2400

0

Cell Resistance (mohm-cm2)

0

Cell Voltage (mV)

CrN, and an inner layer of Cr2 N and/or a Cr–Ni–N phase consistent with the  phase [41–43] ( phase identification based on composition only). The  phase was linked to lower corrosion resistance in PEMFC environments than CrN and Cr2 N, but begins to become unstable above ∼1000.1100 ◦ C [41–43], that is why a 1100 ◦ C nitridation temperature was used for the Ni–Cr base alloys [37]. There was considerable variation in microstructure among the regions analyzed in this sample. At some Cr2 N/alloy interfaces, Al2 O3 was also observed. Occasional porosity in the nitride regions was also evident. Energy dispersive X-ray analysis (EDS) in the STEM indicated that moderate amounts of oxygen were present throughout the nitrided layer. An even more complicated structure was observed in the nitrided G-35姠 region shown in Fig. 4b. The outer surface consisted of complex Cr and Ni base oxides (not fully continuous), likely also containing nitrogen, overlying a thin, inner Cr-nitride layer, indistinguishable between CrN and Cr2 N. A lack of adhesion between the oxide layer and the nitrided layer was evident. At the interface between oxide and nitride, Al2 O3 was also observed. This is in sharp contrast to the nitride layer formed on the G-35姠 coupon shown in Fig. 3a, which was Cr-nitride at the surface. The G-35姠 alloy appeared to have a slightly greater tendency to oxide product formation than G-30䉸 (the alloys were typically nitrided together in the same batch runs), which may be the source of the generally higher ICR values observed for nitrided G-35姠 (Fig. 2). (The G-35姠 coupon shown in Fig. 4b was not measured for ICR). Single-cell fuel cell testing was conducted using nitrided G-35姠 test plates. Performance curves are shown in Fig. 5. Under the RH cycled conditions at 80 ◦ C and 0.2 A/cm2 (Fig. 5a), the behavior was similar to that of graphite plates tested under the same conditions, with no increase in cell resistance due to the plates and no contamination of the MEA with metallic ions (Cr, Fe, Mo, Ni) detected by XRF after 900 h of testing. Visual examination of the test plates also indicated no signs of corrosive attack. The degradation evident in the performance curves (downward trend) is likely due to the RH cycling and subsequent damage to the MEA. Relatively lower performance was observed with a second set of nitrided G-35姠 test plates over the course of a 2700 h test at 60 ◦ C and 0.4 A/cm2 (Fig. 5b). The initial resistance measured between the nitrided G-35姠 test plates was 100 m-cm2 , which was significantly higher than the 60 m-cm2 resistance typically observed for graphite plates tested under the same conditions. Performance declined approximately 25 × 10−6 V/h over the course of the test, about five times higher than a previous test using a nitrided Ni–50Cr and graphite plates. Post-test visual examination indicated significant corrosive attack of the cathode-side nitrided G-35姠 test plate. Preliminary SEM analysis of the anode and cathode test plates revealed extensive Ni, Cr, and Mo-rich oxide products intermixed in the Cr-nitride surface, such that the as-nitrided surface structure on these plates was closer to that shown in Fig. 4b than in Fig 3a, i.e. this particular nitriding run did not yield a high-quality nitrided surface. This is consistent with the relatively high initial resistance observed for these plates. Qualitatively, the extent of oxide products appeared greater in the

3783

Time (h)

Fig. 5. Single-cell fuel cell test data for G-35 anode and cathode plates nitrided for 8 h at 1100 ◦ C in flowing, purified N2 –4H2 : (a) 40%/100% RH cycle test at 80 ◦ C and 0.2 A/cm2 (the performance dips correspond to the cycling in the RH); (b) fully saturated test at 60 ◦ C and 0.4 A/cm2 . Note that a clamping adjustment was made at ∼1820 h, which decreased cell resistance.

lower-performing set of nitrided G-35姠 test plates; however, oxides were also definitely present in the nitrided G-35姠 test plates run at 80 ◦ C and 0.2 A/cm2 with cycling RH, which was anticipated to be the more aggressive test condition but showed excellent behavior. Further analysis of these test plates is planned. 3.2. Nitrided commercial Fe–Cr base alloys Initial efforts to form continuous Cr-nitride surface layers on Fe–Cr base stainless steel alloys were not successful [38,40]. On nitridation at 1100 ◦ C for 24 h, type 349TM (Fe–23.0Cr–14.5Ni–1.5Mn–1.4Si–0.13N–0.40Nb–0.05C wt%) and type 446 MOD-1 stainless steel (Fe–28.37Cr–3.50Mo– 2.96Ni–0.43Mn–0.42Si–0.75(Ti + Nb)–0.03C wt%) formed a discontinuous Cr-nitride surface, with extensive internal nitridation. However, nitridation of type 446 MOD-1 stainless steel at 1100 ◦ C for 2 h in static N2 did reduce ICR values by over an order of magnitude and moderately improve corrosion resistance compared to the untreated alloy [40]. Under these nitridation conditions, the mass uptake was small, generally less than 0.5 mg/cm2 , and the surface appeared tinted. A dense Cr-nitride surface layer was not formed, rather a complex region rich in Cr-oxides, Cr-nitrides, Ti-nitrides, Al-oxides, Fe metal and Cr metal was formed (Si was not detected in the surface), with the level of oxygen detected at the surface

3784

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

(a)

100

60 Cr

Fe 80

40

Atomic concentration, %

Atomic concentration, %

50

N

30 20 O

40

N Ti O

20

Fe

10

60

Cr

Ti 0

0 0 (b)

5

10

15

20

0

25

Sputtering time, min

(c)

5

10

15

20

25

Sputtering time, min

Fig. 6. Surface of AL29-4C after nitridation for 2 h at 1100 ◦ C in flowing, purified N2 –4H2 : (a) SEM secondary electron surface image; (b) AES data for surface particles in (a); (c) AES data for surface matrix in (a). The estimated sputtering rate was ∼50 nm/ min.

significantly higher than the levels of nitrogen [40]. Effectively, nitridation modified the native passive oxide surface of the alloy in a beneficial manner, termed a nitrogen modified passive oxide layer (referred to in the paper as NMPOL for conciseness), with either nitrogen doping of the oxide or local regions of connected nitride phase exposed at the surface resulting in the low ICR values. To further investigate this effect, a series of nitridation experiments were performed on AL29-4C䉸 , a high-Cr ferritic stainless steel similar in composition to type 446 MOD-1 (AL29-4C䉸 is slightly higher in Cr and Mo). Two conditions were selected, 2 h at 1100 ◦ C in flowing, purified N2 –4H2 , in an attempt to minimize the level of oxygen in the nitriding environment, and 800 ◦ C for 24 h followed by 900 ◦ C for 17 h in static N2 –4H2 (no gas purification) to enhance oxide formation in the nitriding environment. Oxidation is enhanced because the level of oxygen needed to oxidize rather than nitride Cr decreases with decreasing temperature; the initial exposure at 800 ◦ C was designed to favor oxidation over nitridation, with the subsequent increase in temperature to 900 ◦ C to shift conditions to favor nitridation (see below). This treatment was also conducted in a sealed environment with a fixed volume of

gas (static condition), such that the oxygen impurity level in the environment would be reduced as oxygen was consumed by oxide formation, which would also be expected to result in a transition to favoring nitride formation over oxide during the course of the treatment. The 1100 ◦ C nitrided AL29-4C䉸 exhibited high mass uptake, 2.7 mg/cm2 , which was the result of extensive internal nitridation in addition to isolated, discontinuous areas of external Cr-nitride formation. SEM imaging of the surface (Fig. 6a) indicated isolated Cr-rich surface particles in a matrix in which the initial 600 grit surface polish lines were still evident (indicative of little external layer formation in these regions). AES depth profiles (Fig. 6b) indicated that the Crrich phase was Cr–(35–40)N–10Fe–(1–2)Ti at%, with a small amount of oxygen present at the very outer surface, i.e. consistent with CrN. AES analysis of the matrix phase at the surface (Fig. 6c) indicated it was Fe-rich, with a Cr level of less than 20 at% Cr, which is depleted relative to the baseline Cr levels of this alloy. This is consistent with extensive internal Crnitride formation in this alloy. At the outer surface, high levels of nitrogen, with smaller levels of titanium and oxygen were also detected, although on profiling the nitrogen and oxygen

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

3785

700

600

ICR (mohm-cm2)

500 AL29-4C Metal

400

300

200 1Nit

AL29-4C

2Nit

100

AL29-4C

(a) 0 0

80

50

100

150

200

2

Compaction Pressure (N/cm )

Fe 70

Atomic concentration, %

O

Cr

60

Fig. 8. ICR data. Nit 1: 2 h at 1100 ◦ C in flowing, purified N2 –4H2 . Nit 2: 24 h at 800 ◦ C and 17 h at 900 ◦ C in static N2 –4H2 (Nit. = nitrided).

50 N 40 30 20 10

Ti

0 0 (b)

10

20

30 40 Sputtering time, min

50

60

70

Fig. 7. Surface of AL29-4C after nitridation for 24 h at 800 ◦ C and 17 h at 900 ◦ C in static N2 –4H2 : (a) SEM secondary electron surface image; (b) AES surface data. The estimated sputtering rate was ∼50 nm/ min.

signal reduced to zero after only a few minutes of sputtering, which is an indication of a very thin nitrided layer. In contrast to the 1100 ◦ C nitridation, the 800 ◦ C-24 h/ 900 ◦ C-17 h treated sample exhibited a relatively fine grained, homogeneous surface (Fig. 7a). Mass uptake as a result of the nitridation treatment was only ∼0.6 mg/cm2 , consistent with a thin surface layer and a minor amount of internal nitridation. AES profiles (Fig. 7b) indicated that the surface consisted solely of Cr and N, of composition range (∼40 at% N) consistent with that of the CrN phase. A small amount of Ti (1 at% range) was also detected at the surface, with Fe not detected. On sputtering, the nitrogen signal gradually decreased, and an oxygen signal ensued (Fig. 7b). The depth profile was therefore indicative of an exclusive Cr-rich surface nitride, overlying a Cr-rich oxide layer. This surface is significantly different than the NMPOL surface formed on type 446 stainless steel, which

contained significant quantities of Fe and O. Rather, it is much closer to the desired continuous or semi-continuous Cr-nitride surfaces formed on nitrided Ni–50Cr [35–37] and the nitrided G-30䉸 and G-35姠 alloys, respectively. ICR values for the two surface treatments are shown in Fig. 8. Untreated AL29-4C䉸 metal exhibited ICR values greater than 100 m-cm2 in the 100.200 N/cm2 contact pressure range of interest for PEMFC stacks, which is consistent with typical ICR values reported for untreated stainless steel alloys [27–29]. Both the nitridation treatments significantly lowered ICR, with the 800 ◦ C-24 h/900 ◦ C-17 h treated surface reaching the goal ICR value of 10 m-cm2 at compaction pressures of only 50–100 N/cm2 and the 1100 ◦ C treated surface at levels of 150.200 N/cm2 . The higher ICR values of the 1100 ◦ C treatment likely reflect the lower volume fraction of conductive CrN particles on the surface, as well as a qualitatively higher degree of surface roughness (Fig. 6a). Polarization data in 1 M sulfuric acid +2 ppm F− at 70 ◦ C is shown in Fig. 9, under aerated (9a) and hydrogen-purged (9b) conditions. The 1100 ◦ C treated surface exhibited relatively poor corrosion resistance, consistent with a nonprotective, discontinuous CrN surface, with the surrounding matrix made vulnerable due to Fe-enrichment/Cr depletion from internal Cr-nitride formation. The 800 ◦ C-24 h/900 ◦ C-17 h treated AL29-4C䉸 exhibited excellent corrosion resistance under these aggressive conditions, consistent with a protective CrN surface layer. Fig. 10 shows ICR data for an AL29-4C䉸 coupon nitrided at 900 ◦ C for 24 h in N2 –4H2 (0.24 mg/cm2 uptake) before and after a 7.5 h hold in aerated 1 M sulfuric acid +2 ppm F− at 70 ◦ C and +0.84 V to simulate aggressive PEMFC cathodic conditions. The anodic current during polarization was only ∼0.3 × 10−6 A/cm2 . The ICR increased only slightly, similar to that observed for nitrided G-35姠 (Fig. 2b). It should be

3786

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

350 Nit AL29-4C

300 2x ICR (mohm-cm2)

Potential (V vs SHE)

1.2

2

0.9

0.6

0.3 1

250 200

Polarized As-Nitrided

150 100 50

Nit AL29-4C

Goal

0 0

0 1E

-8

-6

1E

1E

-2

Current Density (A/cm2)

(a)

2

1.2

Potential (V vs SHE)

1E

-4

0.9

0.6

0.3 AL29-4C

0 1E-8 (b)

1E-6

1E-4

100

150

200

2

Compaction Pressure (N/cm ) Fig. 10. ICR data for AL29-4C nitrided for 24 h at 900 ◦ C in static N2 –4H2 . Polarization for 7.5 h at 0.84 V vs SHE in aerated 70 ◦ C 1 M sulfuric acid + 2 ppm F− . Note that ∼2x ICR is plotted.

Nit AL29-4C

1Nit

50

1E-2

Current Density (A/cm2)

Fig. 9. Polarization data in 70 ◦ C 1 M sulfuric acid + 2 ppm F− : (a) aerated; (b) H2 purged. Nit 1: 2 h at 1100 ◦ C in flowing, purified N2 –4H2 . Nit 2: 24 h at 800 ◦ C and 17 h at 900 ◦ C in static N2 –4H2 (Nit. = nitrided).

noted that there was significant variability from run to run in the mass uptake and surface structure of AL29-4C䉸 nitrided at 800–900 ◦ C in static and flowing/purified N2 –4H2 . In some cases, high mass uptakes (2.5 mg/cm2 range) and extensive internal nitridation occurred and yielded poor corrosion resistance, similar to that observed for the AL29-4C䉸 nitrided at 1100 ◦ C for 2 h in purified N2 –4H2 (Figs. 6 and 9). 3.3. Implications Commercially available, corrosion-resistant Ni–Cr and Fe–Cr base alloys were designed with regards to protective oxide layer formation (among other mechanical property and manufacturability considerations), not to be nitrided to

form continuous, protective Cr-nitride surface layers. Despite this, nitridation of G-30䉸 , G-35姠, and AL29-4C䉸 yielded a promising combination of low ICR values and good corrosion resistance in simulated PEMFC environments. Unlike the model Ni–50Cr alloy, the surface nitride layer formed on all three alloys was closer to semi-continuous Cr-nitride than fully continuous. This is reflected in the microstructures observed for the nitrided Ni–Cr base alloys. For the nitrided AL29-4C䉸 , a definitive assessment of Cr-nitride continuity was not made. Based on the slight increase in ICR on polarization, it is possible that this surface was also semi-continuous, as no increase in ICR was observed for nitrided Ni–50Cr. Fuel cell testing will be needed to determine if such a surface can meet performance and durability goals. Further optimization may also result in a transition to fully continuous nitride surface layers. In the case of the Ni–Cr alloys, oxygen impurities in the nitriding environment caused significant variability from run to run, with oxide formation at the external surface compromising the protectiveness of the surface layer that was formed. This was particularly evident in nitrided G-35姠 test plates evaluated in the single-cell fuel cell testing. More rigorous nitriding protocols than used in the present work are necessary to ensure exclusive nitride formation at the surfaces of these alloys. These include more rapid heating cycles, to minimize the impact of higher Cr-oxide vs Cr-nitride stability at lower temperatures, vacuum bake outs to better eliminate adsorbed moisture in the furnace, and higher levels of hydrogen in the nitriding environment to lower the oxygen activity. A complication is the presence of alloying additions and/or alloy impurities such as Al, Si, and Ti, which are more stable thermodynamically with oxygen than is Cr. As such, it may be difficult to completely eliminate oxide formation. The internal nitridation in these alloys accompanying the external nitride layer formation is also a concern due to the possibility of embrittling the alloy. This

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

is particularly an issue with stamped alloy foil bipolar plates, for which nitridation of 0.1 mm thick G-30䉸 foil at 1100 ◦ C for 6 h in N2 –4H2 was observed to cause embrittlement. Modification of nitridation parameters to limit this effect therefore also needs to be considered. In the case of the ferritic AL29-4C䉸 alloy, oxygen impurities in the nitriding environment appeared to help promote the formation of an external Cr-nitride layer. The initially formed oxide is believed to reduce nitrogen permeation into the alloy, which otherwise would result in extensive internal nitridation, effectively partitioning nitrogen away from the alloy surface [31,44]. The mechanistic reasons behind the differences in oxygen effects on nitridation between this alloy and the Ni–Cr alloys are not readily apparent, and are the subject of ongoing study. A key difference was the formation of Cr, Ni and Mo containing oxides at the surface of some nitrided Ni–Cr alloy samples, whereas Cr-rich oxide, generally free of base metal Fe, was detected underneath the nitride surface layer on AL29-4C䉸 . It is possible that controlled use of oxygen impurities/oxide formation could be used to improve Cr-nitride layer formation on Ni–Cr base alloys, similar to that observed for AL29-4C䉸 in the present work. The dependence on trace levels of oxygen impurities to yield protective Cr-nitride surfaces on AL29-4C䉸 did, however, result in run to run variability. This suggests the use of a preoxidation step during the course of the nitriding treatment, using gases with defined levels of oxygen added to controllably introduce a set amount of oxygen/oxide formation prior to nitridation. (The dual nitridation temperature 800/900 ◦ C treatment relied on unspecified levels of oxygen impurities, which can (and did) vary from one gas cylinder to the next). Efforts devoted to this approach, as well as design of ferritic alloys specifically to be nitrided are in progress [31,44]. Initial findings suggest promising results with a preoxidation step, particularly in developmental Fe–Cr ferritic alloys modified by small amounts of V to aide in the conversion of the initial oxide surface to nitride. The use of preoxidation also significantly reduces internal nitride formation, and thus would ameliorate potential concerns regarding embrittlement of thin foil bipolar plates. 4. Summary (1) Nitridation of the commercial Ni–Cr base alloys G-30䉸 and G-35姠, and the ferritic stainless steel AL29-4C䉸 , resulted in a significant reduction in ICR values, to the range of 10 m-cm2 at compaction pressures of less than 200 N/cm2 . Only slight increases in ICR were observed on polarization under aggressive simulated PEMFC cathodic conditions. Nitridation also yielded corrosion resistant surfaces, with anodic current densities generally less than 1 × 10−6 A/cm2 up to +0.9 V vs SHE under simulated PEMFC cathodic conditions, and from −0.2 to +0.1 V vs SHE under simulated PEMFC anodic conditions. (2) It was possible to form semi-continuous Cr-nitride surface layers on G-30䉸 , G-35姠, and AL29-4C䉸 by nitridation. Variability in the surface structure that formed was observed

3787

from nitriding run to run, and resulted from oxygen impurities in the nitriding environment. The oxygen impurities were generally detrimental for nitridation of G-30䉸 and G-35姠, resulting in oxide formation at the surface under some conditions, but beneficial for the nitridation of AL29-4C䉸 , by aiding growth of the Cr-nitride at the outer surface. (3) Overall, thermal nitridation shows good promise to protect metallic bipolar plates and meet property/performance targets. Acknowledgments The authors thank John Shingledecker, S.J. Pawel, and P.F. Tortorelli for helpful comments in reviewing this manuscript. The authors also thank L. Flower of Haynes International and J. Rakowski of Allegheny Ludlum Inc for technical input and for supplying the alloy material used in this work, and G. Teeter conducting the AES measurements. Funding from the United States Department of Energy Hydrogen, Fuel Cells, and Infrastructure program is gratefully acknowledged for the work performed by ORNL and NREL. Oak Ridge National Laboratory is managed by UT-Battelle, LLC for the US DOE under contract DE-AC05-00OR22725. References [1] Larminie J, Dicks A. Fuel cell systems explained. West Sussex, England: Wiley; 2000. [2] Borup R, Vanderborgh NE. In: Doughty DH, Vyas B, Takamura T, Huff JR, editors, Material research society symposium proceedings, vol. 393. Pittsburgh, PA, USA: Material Research Society; 1995. p. 151. [3] Steele BCH, Heinzel A. Nature 2001;414:345–52. [4] Brandon NP, Skinner S, Steele BCH. Annu Rev Mater Res 2003;33:183. [5] Hermann A, Chaudhuri T, Spagnol P. Int J Hydrogen Energy 2005;30:1297. [6] Besmann TM, Klett JW, Henry JJ, Lara-Curzio E. J Electrochem Soc 2000;147:4083–6. [7] Scholta J, Rohland B, Trapp V, Focken U. J Power Sources 1999;84: 231–4. [8] Wilkinson DP, Lamont GJ, Voss HH, Schwab C. Method of fabricating an embossed fluid flow field plate. US Patent 5,527,363, June 18; 1996. [9] Cunningham N, Guay D, Dodelet JP, Meng AR, Hlil AS, Hay AS. J Electrochem Soc 2002;149:905–11. [10] Lawrance RJ. Low cost bipolar current collector–separator for electrochemical cells. US Patent 4,214,969, July 29; 1980. [11] Balko EN, Lawrance RJ. Carbon fiber reinforced fluorocarbon-graphite bipolar current collector–separator. US Patent 4,339,322, July 13; 1982. [12] Busick DN, Wilson MS. Composite bipolar plates for fuel cells. In: Proton conducting membrane fuel cells II, vol. 98-27. Boston, Ma: The Electrochemical Society; 1998. p. 435–45. [13] Wilson MS, Busick DN. Composite bipolar plate for electrochemical cells. US Patent 6,248,467, June 19; 2001. [14] Wang JH, Baird DG, Mcgrath JE. J Power Sources 2005;150:110. [15] Abd Elhamid MH, Blunk RH, Lisi DJ, Mikhail MY. Polymer composite. US Patent Application 20040062974, June 26; 2003. [16] Hentall PL, Lakeman JB, Mepsted GO, Adcock PL, More JM. J Power Sources 1999;80:235–41. [17] Pozio A, Silva RF, De Francesco M, Giorgia L. Electrochim Acta 2003;48(11):1543–9. [18] Makkus RC, Janssen AHH, de Bruijn FA, Mallant RKAM. J Power Sources 2000;86:274. [19] Davies DP, Adcock PL, Turpin M, Rowen SJ. J Appl Electrochem 2000;30:101–5.

3788

M.P. Brady et al. / International Journal of Hydrogen Energy 32 (2007) 3778 – 3788

[20] Tarutani Y, Doi T, Seki A, Fukuta S. Stainless steel product for producing polymer electrode fuel cell, U.S. Patent 6,379,476, April 30; 2002. [21] Joseph S, McClure JC, Chianelli R, Pich P, Sebastian PJ. Int J Hydrogen Energy 2005;22(12):1339. [22] Li M, Luo S, Zeng C, Shen J, Lin H, Cao C. Corros Sci 2004;46: 1369–80. [23] Lee SJ, Huang CH, Chen YP. J Mater Process Tech 2003;140:688–93. [24] Aukland N, Boudina A, Eddy DS, Mantese JV, Thompson MP, Wang SS. J Mater Res 2004;19(6):1723. [25] Ma CL, Warthesen S, Shores DA. J New Mater Electrochem Sys 2000;3:221. [26] Wind J, Spah R, Kaiser W, Bohm G. J Power Sources 2002;105:256. [27] Wang H, Sweikart M, Turner JA. J Power Sources 2003;115:243–51. [28] Wang H, Turner JA. J Power Sources 2004;128:193. [29] Wang H, Teeter G, Turner JA. J Electrochem Soc 2005;153(3):B99. [30] Fan Q, Chervinko JR, Onischak M, Marianowski LG. Low cost metal bipolar plates and current collectors for polymer electrolyte membrane fuel cells. US Patent application number 20020172849, April 6; 2001. [31] Brady MP, Yang B, Wang H, Turner JA, More KL, Wilson M. et al. JOM-J Metals Miner Mater Soc 2006;8:50–7. [32] Weil KS, Xia G, Yang ZG, Kim JY. Development of a niobium clad PEM fuel cell bipolar plate material, Int J Hydrogen Energy, in press, this issue, doi:10.1016/j.ijhydene.2006.08.041.

[33] Abdelhamid MH, Blunk RH, Vyas G. Enhanced stability bipolar plate. US Patent Application 20060019142, Jan 26; 2006. [34] Brandl W, Gendig C. Thin Solid Films 1996;343:290–1. [35] Brady MP, Weisbrod K, Zawodzinski C, Paulauskas I, Buchanan RA, Walker LR. Electrochem Solid-State Lett 2002;5:A245–7. [36] Brady MP, Weisbrod K, Paulauskas I, Buchanan RA, More KL, Wang H. et al. Scr Mat 2004;50(7):10–7. [37] Paulauskas I, Brady MP, Meyer III HM, Buchanan RA, Walker LR. Corrosion behavior of CrN, Cr2 N and PI phase surfaces formed on nitrided Ni–50Cr with application to proton exchange membrane fuel cell bipolar plates. Corros Sci, in press, doi:10.1016/j.corsci.2005.10.019. [38] Wang H, Brady MP, Teeter G, Turner JA. J Power Sources 2004;138: 86–93. [39] Rubly RP, Douglass DL. Oxid Met 1991;35(3–4):259–78. [40] Wang H, Brady MP, More KL, Meyer HH, Turner JA. J Power Sources 2004;138:79–85. [41] Ono N, Kajihara M, Kikuchi M. Metal Trans A 1992;23:1389–93. [42] Kodentsov AA, Gulpen JH, Cserhati C, Kivilahti JK, VanLoo FJJ. Metal Mater Trans A 1996;27:59–69. [43] Krupp U, Chang SY, Christ H-J. Z Metallk 2000;91:1006–12. [44] Yang B, Brady MP, Wang H, More KL, Tortorelli PF, Young DJ, Payzant EA, Turner JA, submitted for publication.