Effects of aging and nitridation on microstructure and mechanical properties of austenitic stainless steel

Understanding the effects of in-service microstructural changes in metal alloys on the mechanical performance and remaining life of equipment is a critical part of integrity management in industrial plants. However, the effects of processes such as carburization or nitridation are not accounted for in industry-standard life assessment methodologies such as those provided in the API 579-1/ASME FFS-1: Fitness-For-Service standard. This is problematic for the austenitic stainless steel Alloy 800H, which typically operates in the creep regime and has been reported to suffer nitriding during high-temperature service in air. Despite this being one of the most common service environments for 800H, the impact of nitriding on in-service performance and implications for remaining life assessment have not been well-studied for this alloy. In this work, we characterize the microstructures of as-received, aged, and nitrided Alloy 800H tube material, and correlate observations with room-temperature tensile properties and high-temperature creep behavior. We show that while the creep properties of aged 800H material can be captured by the widely-used MPC Project Omega creep model and API 579-1 Omega properties for 800H, nitrided material properties fall outside of the expected bounds, and therefore remaining life cannot be reliably predicted using this method without experimental data.


Introduction
For metal alloys used in industrial applications, it is often critical that their mechanical properties are well-characterized and the effects of any microstructural changes that may occur in service are understood. This information is used in component design to ensure an acceptable lifetime will be achieved, but also for condition monitoring during service to determine when replacements should be undertaken. Standards such as API 579-1/ASME FFS-1: Fitness-For-Service (API 579-1) [1] are heavily relied on in industry for providing guidance around when it is appropriate to return a component to service, repair it, or replace it. However, the outcome of an API 579-1 assessment is strongly influenced by the mechanical property information available for the material of interest.
Relevant properties are provided in API 579-1 and other standards such as API 530: Calculation of Heater Tube Thickness in Petroleum Refineries (API 530) [2] for a wide range of industrial alloys at service-relevant temperatures. However, the alloys included in these standards are limited to those considered to have properties that demonstrate little or no change over time under typical service conditions. An example of a commonly used group of alloys omitted from these standards are the high-temperature HP alloys used for steam reformer catalyst tubes. These precipitationstrengthened alloys are known to undergo aging in service that results in a gradual reduction in creep strength [3]. Service conditions that lead to uptake of species from the environment (e.g., carburization, nitridation, sulfurization) are also not accounted for in the properties provided in API 579-1 and API 530. Material performance information can be found elsewhere in the literature for some alloy-environment combinations relevant to real service conditions, but there are still many that are yet to be studied.
Alloy 800H is one of the alloys included in API 579-1. It is an austenitic stainless steel often used for high-temperature service where good creep strength and corrosion resistance are necessary, but the expense of using nickel-based alloys is not justified [4]. Classified by manufacturers as a solid-solution strengthened alloy [4,5], commercial heats usually contain a low volume fraction of Ti(C,N) and M 23 C 6 precipitates [4], and it is known to develop a c 0 phase at temperatures below 700°C [6,7]. 800H also forms an external, primarily Cr 2 O 3 scale when exposed to oxidizing conditions. A common application for 800H is the process flow outlet tubes on steam reformers, referred to as ''pigtails''. Pigtails are typically exposed to air at the external surface and syngas (a mixture of H 2 , CO, CO 2 , and some remnant CH 4 and steam) at the internal surface, and usually operate at temperatures around 800-900°C [8]. These conditions pose some risk of carburization at the internal surface and nitridation at the external surface of the tubes, but it is generally assumed that the Cr 2 O 3 scale provides sufficient protection against these microstructure-altering processes.
Several metallurgical studies of ex-service Alloy 800H pigtails suggest otherwise. Hoffman and Lai [8] observed aluminum and chromium nitrides in both creep-damaged material and undamaged material near the external (air-exposed) surface of pigtails that had been in service for 6 to 10 years. Smillie et al. [9] reported extensive AlN formation around surfacebreaking creep cracks and in regions near the external surface of pigtails that had seen 7 years' service. Spyrou et al. [10] also found AlN precipitates near the external surface of pigtails that had been in service 7 years. These studies indicate that internal nitridation can be an issue for Alloy 800H exposed to air, albeit one that may not become significant until later in life. However, as a component approaches end-oflife and the need for accurate fitness-for-service and remaining life assessment increases, it becomes more critical to understand the effect that issues such as this have on mechanical performance.
API 579-1 provides guidance for remaining life assessment of components operating in the creep regime using either the Larson-Miller or MPC Project Omega models. The Larson-Miller model defines a so-called Larson-Miller parameter (LMP) for a given stress that allows rupture life to be predicted for any given temperature, according to Eq. (1): where T is in Kelvin, C is the Larson-Miller constant for the material of interest, and t r is the time to rupture in hours. The Larson-Miller model is also used in API 530 for design of high-temperature tube components, where a design lifetime of 100000 h (11.4 years) is standard unless specified otherwise. Both standards provide both mean and lower-bound values for C, which allows for some level of variability in material properties.
The MPC Project Omega model calculates rupture time as a function of the strain rate parameter _ co and the multi-axial Omega parameter X m that describes creep ductility, Eq. (2) [1,11]: dimensionless adjustment parameters used to characterize strain rate and creep ductility, D sr and D cd , respectively, are expected to vary within certain bounds. Again, this allows the model to capture some level of variability in material properties.
Specific values of C and D sr and D cd can be obtained by fitting the respective models to creep test data, but assessments where no experimental data is available are often conducted assuming lower bound values from API 579-1 for conservatism. It is possible for experimental values to fall outside of the expected bounds; this corresponds to behavior outside of the scatter band upon which the API 579-1 models are based. Currently, it is not clear whether the behavior of nitrided Alloy 800H material lies within the material property bounds provided in API 579-1, and therefore whether component remaining life can be accurately assessed without experimental mechanical property data. Review of the literature further suggests that there is not yet any published data for material properties of nitrided Alloy 800H material that can be used to investigate this question. This is the focus of the present work. We subjected samples of Alloy 800H pigtail material to a nitridation heat treatment that yielded material with through-thickness nitride precipitation. Hardness, room-temperature tensile properties, and creep performance at 1000°C of this nitrided material were evaluated and compared to that of material as-received from the supplier and material subjected to an aging heat treatment in argon gas. We also characterized the microstructure of the pigtail material in all three conditions using optical, scanning electron, and transmission electron microscopy, and correlated observations to the measured mechanical properties. We demonstrate how both aging and nitridation can alter the mechanical performance of Alloy 800H and discuss the implications for life prediction and inservice fitness-for-service assessments.

Materials and methods
The raw material used in this study was commercially manufactured Alloy 800H NPS 1-1/4'' Sch. 160 seamless pipe (nominal outside diameter 42.16 mm and wall thickness 6.35 mm), supplied in a solution annealed and water quenched condition. Four samples of length approximately 150 mm were used, all from the same original heat. The average sample composition measured using optical emission spectroscopy is given in Table 1 along with the ASTM B407-08a [12] specification for Alloy 800H/UNS N08810.
Specimens in three different material conditions were prepared: as-received (AR condition), aged in 99.99% argon for 800 h at 1000°C and 1 atm. pressure (AA condition), and nitrided in a 95% N 2 ? 5% H 2 gas atmosphere for 800 h at 1000°C and 1 atm. pressure (NI condition). The heat treatment conditions for the NI specimens were chosen to result in through-thickness AlN precipitation, determined from our previous work on nitridation kinetics in Alloy 800H [13].
Aging and nitridation heat treatments for the AA and NI specimens, respectively, were based on the procedures described in ISO 21608 [14]. A horizontal Lindberg/Blue M 55000 series hinged furnace with an alumina furnace tube sealed from the atmosphere was used to conduct the heat treatments. Coupons with dimensions 82 9 15 9 4 mm were placed in individual tubular alumina crucibles inside the main furnace tube and an independent N-type thermocouple was used to monitor coupon temperature. Readings from this thermocouple were used to set the furnace temperature for each heat treatment such that coupon temperature was maintained at 1000°C. Prior to beginning each heat treatment, the furnace tube with the coupons inside was subjected to two cycles of evacuation and backfilling with the test gas to remove oxygen from the system. To maintain a low oxygen partial pressure during heat treatments, titanium sponge was also placed inside the furnace tube to act as an oxygen getter.
Following heat treatment, samples of AR, AA and NI material were prepared for microstructural characterization using optical and scanning electron microscopy (SEM), including electron backscatter diffraction (EBSD) and energy dispersive spectroscopy (EDS) for phase identification. The preparation methods and analysis undertaken were as described in [13]. Samples of AR and NI material were also prepared for and analyzed using transmission electron microscopy (TEM), including selected area diffraction (SAD), as described in [15]. In addition, hardness traverses were undertaken on at least three mounted cross-sectional samples in each material condition using a Leco M-400-H1 Micro Vickers hardness indenter and a 1 kg load.
Three tensile specimens and four creep specimens were prepared for each material condition. Geometries were adapted from miniature specimen designs in ASTM E8 [16] to accommodate raw material geometry and testing equipment requirements and are shown in Supplementary Figures S1 and S2. Prior to testing, the large faces of all specimens were ground with SiC paper up to 600-grit and polished to a 9 lm finish using Buehler MetaDi TM diamond suspension and Engis HyprezÒ diamond compound on a Buehler MasterTex TM pad.
Tensile tests to failure were conducted at room temperature on an MTS 810 Material Testing System with a 100 kN load unit, and a 25 mm gage MTS extensometer was used to record strain. The displacement rate for all tests was 1 mm min -1 . The cross-sectional area of the gage length of each specimen was measured in three locations before testing, and the minimum area was used to calculate stress.
All creep tests were carried out in air at 1000°C under a nominal stress of 8 MPa using a custommade constant-load test rig (see [17] for design details). Test conditions were selected based on the creep deformation mechanism map (CDMM) for Alloy 800H developed by Beardsley et al. [18], and were intended to replicate the in-service deformation mechanism for steam reformer pigtail tubes as closely as possible while still providing results in a reasonable timeframe. All four specimens of one material condition were tested simultaneously. The raw data comprised total elongation of each individual specimen and furnace temperature at the center point of the gage length of each specimen at one-minute intervals. The cross-sectional area of the gage length of each specimen was measured in three locations before testing, and the minimum area was used to calculate stress from the applied load. Elongation was converted to strain by dividing by the initial 30 mm gage length. Tests were stopped once all specimens had either reached 5% strain or were otherwise deemed to have achieved a minimum in strain rate. Following testing, cross sections from selected creep specimens were prepared as described in [13] for examination using optical and SEM methods. Hardness traverses were recorded from these specimens using a Leco M-400-H1 Micro Vickers hardness indenter and a 1 kg load.

AR material
The microstructure of the AR material studied has been described in detail in our previous works, see [13,15], but a brief review of the key aspects is provided here. The grain structure of the austenite (c) matrix was equiaxed with a large proportion of annealing twins. The average twin-excluded grain diameter was approximately 100 lm. EBSD analysis of grain boundary character distribution and classification of boundaries according to coincident site lattice nomenclature indicated that approximately 50% (length fraction) of boundaries were random high-angle boundaries (RHABs), 45% were R3 boundaries, and 5% were R3 n boundaries where n C 2.
Dendritic M 23 C 6 precipitates \ 200 nm thick were present on RHABs in the AR material. These had a high length/width to thickness ratio, and were similar in appearance to cementite formations on austenite grain boundaries characterized by Kral and Spanos [19,20]. Although not common, dendritic M 23 C 6 precipitates have been reported by others [21,22]. SAD showed that these precipitates typically displayed a cube-cube orientation relationship (OR) with the c matrix in one of the adjacent grains; this is known to occur when M 23 C 6 precipitates on c grain boundaries [23]. Intragranular Ti(C,N) precipitates were uniformly distributed throughout the microstructure, and were classified into two size All composition values are in wt% categories: large precipitates, often square or rectangular in cross section, with side length on the order of 10 lm; and small precipitates with a rhomboid cross section and side length on the order of 0.1-1 lm. All of the latter that were analyzed using SAD also displayed a cube-cube OR with the c matrix. SEM images of the general AR microstructure and precipitates are shown in Fig. 1a, c, e.

AA material
Compared to AR material, the primary difference in the AA material microstructure was the size and distribution of precipitates. In AA material, M 23 C 6 precipitates on RHABs were coarser and more blocky, and did not cover the full length of the grain boundaries. There was no discernible change in the large size class of Ti(C,N) precipitates, but the small Ti(C,N) precipitates that remained were generally larger in size and fewer in number than in the AR material. Examples of these changes in precipitate size and distribution are shown in Fig. 1. There was no systematic increase in grain size following the aging heat treatment, and no appreciable change in grain boundary character distribution.

NI material
The microstructure of nitrided Alloy 800H has also been described in detail in our previous works [13,15]. However, these studies focused on material that had been exposed to a nitriding atmosphere for times less than 800 h at temperatures 800-1000°C. A brief description of microstructure is provided here that is more specific to the NI specimens studied in this work (nitrided at 1000°C for 800 h). The grain structure of NI material was again similar to that of AR material, with no discernible change in grain size or grain boundary character distribution. The principal difference between AR and NI material was the precipitate phases. Three nitride precipitate phases (CrN, Cr 2 N, and AlN) were observed in NI material, along with secondary precipitation of small, intragranular M 23 C 6 and Ti(C,N) precipitates on the order of 1 lm or less in size. The nitride phases appeared as three uniformly thick precipitation fronts extending from the free surface of each specimen inwards toward the center. The location of the precipitation front for each nitride M x N was designated x MxN , as measured from the free surface x = 0, CrN . In our previous work [15] we found that the secondary M 23 C 6 and Ti(C,N) precipitation in Alloy 800H also appeared as uniform fronts preceding the AlN front (i.e., x M23C6 [ x AlN and x Ti(C,N) [ x AlN ). A schematic illustrating the different regions within the precipitation zone and the relative locations of all precipitation fronts is provided in Fig. 2a.
The nitridation heat treatment was designed to achieve through-thickness AlN penetration (i.e., x AlN & 2 mm). This was achieved in all NI specimens, meaning that distinct AlN and secondary M 23 C 6 and Ti(C,N) fronts were not observed. Cr 2 N and CrN precipitates did not penetrate through the full thickness: x Cr2N = 848-912 lm and x CrN \ 100 lm in all NI specimens. Note that the small penetration depth of CrN meant that these precipitates were completely removed during machining of mechanical test specimens. They therefore had no influence on the observed mechanical properties and are omitted from further discussion.
A range of precipitate morphologies were observed in NI material, see Fig. 2 for examples. In the region x \ x Cr2N , intragranular AlN and Cr 2 N precipitates were typically acicular in cross section and grainboundary precipitates were blocky. Beyond the Cr 2 N front, intragranular AlN precipitates were large (on the order of 10 lm in diameter) and became increasingly blocky and less acicular. Intragranular Ti(C,N) precipitates were also visible in this region, both evenly dispersed through the matrix and more densely collected in string-like formations. Dendritic M 23 C 6 precipitates were also observed on RHABs. These formations were more substantial than those in the AR material, but not as coarse as those in AA material.

Hardness testing
Hardness traverses from specimens in each material condition are shown in Fig. 3. The hardness of AA specimens was uniform. The AR sample hardness was also uniform apart from near surface measurements that had higher values. This surface hardening is likely from residual cold work due to machining. The average hardness of AR specimens was 165 HV1, compared to 156 HV1 for AA specimens. NI specimens showed a decrease in hardness from approximately 305 HV1 at the free surface to a minimum of 177 HV1 in the center of specimens (2000 lm from the free surface), and a larger variation between measurements compared to AR and AA specimens.

Tensile testing
The stress-strain tensile curves from each specimen are shown in Fig. 4. The average mechanical properties (elastic modulus, E; 0.2% offset yield strength, r YS ; ultimate tensile strength, r UTS ; and plastic strain at failure, e p ) derived from these curves are reported in Table 2. Images of the different fracture surfaces for each material condition are also provided in Supplementary Figure S3.
There were no significant differences between the properties of AR and AA material except for r UTS , where AR material demonstrated a higher r UTS than AA material. E was approximately the same for all conditions. In contrast, there was a clear difference between the other tensile properties of AR and NI material, and AA and NI material. The NI material showed a higher r YS and lower r UTS and e p compared to both AR and AA material.
The relatively low r UTS of the NI material was unexpected given the high r YS . However, a large number of cracks appeared on the surface of each NI specimen close to the yield point during testing. Posttest microstructural examination revealed that these cracks were intergranular and appeared to propagate only through grain-boundary Cr 2 N precipitates (rather than along the precipitate-matrix interface, see Fig. 5), arresting at x Cr2N . These cracks therefore would have reduced the effective cross-sectional area of the specimens. To give an estimate of r UTS with this reduction in area taken into account, the original values were adjusted assuming the cross-sectional area at the point of maximum loading was equal to the area of the specimen that contained no Cr 2 N precipitates (rather than the total initial cross-sectional area). This gave an adjusted average r UTS of 595 MPa, which is very similar to r UTS for AR material.

Mechanical properties
The raw strain vs. time creep test data for each specimen is plotted in Fig. 6. There was some variation in creep strain rates across the four specimens tested for each material condition, despite these being tested under the same nominal conditions, but there was still a marked difference between each material condition. To quantify the differences, a hyperbolicsine power-law curve of the form shown in Eq. (3) was fitted to the strain vs. time curve for each creep Figure 4 Stress-strain curves from three room-temperature tensile tests of a AR, b AA, and c NI specimens.
A parametric plot of minimum strain rate ( _ min ) against time to minimum strain rate (t min ) is shown in Fig. 7. More detailed test information, including _ min and t min values, test temperature and stress, and specimen average grain size, is provided in Supplementary Table S2. It is apparent that _ min in all NI specimens was close to one order of magnitude lower than that of AR specimens, and at least one order of magnitude lower than that of AA specimens. For AA specimens, _ min was typically higher than that of AR specimens, although the difference between these two conditions was less than an order of magnitude. t min was similar for AR and AA specimens, but t min for NI specimens was an order of magnitude greater than for AR and AA specimens.

Microstructures
Following creep testing, at least one specimen of each material condition was sectioned and the microstructure examined. Given the wide range of strains at test termination for both the AR and AA conditions (see Fig. 6), the specimens with the maximum and minimum strain at test termination were examined for both of these material conditions. The NI specimens showed a much smaller range of strains, so only one NI specimen was examined in detail.  All specimens had developed an external oxide scale of varying thickness, identified as primarily Cr 2 O 3 with minor amounts of a Mn-rich spinel in some areas. Immediately beneath the external scale, SiO 2 precipitates had formed. Further from the free surface, Al 2 O 3 had precipitated both on grain boundaries and in the matrix immediately adjacent to the grain boundaries. The depth of Al 2 O 3 penetration varied between the different material conditions, which was attributed to the different test times, but was approximately 150 lm for AR specimens, 120 lm for AA specimens, and 200 lm in the NI specimen examined.
Beyond these near-surface regions, the microstructure of the minimum-strain AR and AA specimens (AR12-05 and AA12-03, respectively) was very similar to that of the AA material prior to creep testing, with one exception. For both material conditions, a band of fine (\ 1 lm diameter) Ti(C,N) precipitates had formed adjacent to the Al 2 O 3 region. These Ti(C,N) precipitates were more densely distributed than those in the center of the specimens. In AR12-05, this band occupied the region 120-400 lm from the free surface, and in AA12-03 100-700 lm from the free surface.
The most significant microstructural changes occurred in the maximum-strain AR and AA specimens (AR09-06 and AA09-04, respectively). A large number of macro-cracks had developed in AR09-06, some of which were surrounded by a halo of nitride precipitates, see Fig. 8a. Both AlN and Cr 2 N precipitates had formed, but the AlN front reached a greater distance from the crack surfaces than the Cr 2 N. An increased density of fine intragranular Ti(C,N) was also observed beyond the edge of the AlN front, but no intragranular M 23 C 6 was detected.
Nitride precipitates had also formed in AA09-04. The key difference in this specimen was that only AlN had formed, and it was located in a band adjacent to the region of Al 2 O 3 near the surface of the specimen, Fig. 8b. There was some macro-cracking in AA09-04, but cracks were significantly smaller and more sparsely distributed than in AR09-06, and none were surrounded by the nitride halos observed in that specimen. Nitrides and creep damage aside, the remainder of the AR09-06 and AA09-04 microstructures were very similar to those of AR12-05 and AA12-03, respectively.
NI specimens had also undergone some degree of microstructural change during creep testing. The surface and near-surface regions were similar to those of AR and AA specimens, comprising an external Cr 2 O 3 scale, a thin sub-surface layer of SiO 2 precipitates, and a mixture of grain boundary and fine intragranular Al 2 O 3 precipitates penetrating to a depth of approximately 200 lm. No nitrides were observed within the Al 2 O 3 zone.
Beyond this near-surface region the original nitride phases were still present but had undergone some changes. Prior to testing, acicular and blocky Cr 2 N precipitates were present intragranularly and on grain boundaries, respectively, and were confined to the region within 480 lm of the free surface. After creep testing, coarse Cr 2 N precipitates were distributed on grain boundaries throughout the specimen, including in the center region where they had previously been absent, Fig. 9a, b. There were very few intragranular Cr 2 N precipitates remaining, and the grain boundary precipitates were almost exclusively confined to RHABs, Fig. 9c. In contrast, there was no clear change in the morphology or distribution of AlN precipitates. Hardness traverses taken from the grip section of NI specimens following testing showed that the through-thickness hardness profile of the NI specimens had become uniform (i.e., there was no longer a gradient in hardness as shown in Fig. 3), with an average value of 213 HV1.

Discussion Hardness
The slightly lower hardness of the AA material compared to the AR material (Fig. 3) was consistent with the observed microstructures. The fine, closely spaced Ti(C,N) precipitates in AR material (Fig. 1e) are expected to provide a greater precipitationstrengthening effect than the coarser and more sparsely distributed Ti(C,N) in AA material (Fig. 1f). However, the average difference in hardness between the two material conditions was only approximately 10 HV1, indicating that the influence of the Ti(C,N) precipitates on room-temperature deformation behavior was not significant.
The higher hardness of NI material (Fig. 3) is attributed to both the strengthening effect of the nitride precipitates, and solid-solution strengthening provided by additional nitrogen in the c matrix. Note that the nitriding heat treatment was only intended to achieve through-thickness AlN and not a uniform microstructure, so the alloy nitrogen concentration (and therefore nitride volume fraction and matrix nitrogen concentration) is expected to decrease with increasing distance from the free surface. This is consistent with the observed decrease in hardness with increasing distance from the free surface.

Tensile properties
The room-temperature tensile properties obtained for AR material ( obtained total elongation values in the range 0.5-0.6, but neither reported e p or E.
Manufacturer datasheets for Alloy 800H typically put E at 20°C at around 195 GPa [4,5], and Hammond et al. reported a value of 194.3 GPa at 24°C [26]. However, these figures were obtained from dynamic test methods, rather than the static method used in this work. Uncertainty in E obtained from dynamic testing has been reported at 2% compared to 12% reported for static tests [27]. The E values obtained in the present work for AR material are within 12% of the dynamic E values reported by others [4,5,26] so are considered reasonable given the use of a static test method.
The differences in tensile properties obtained for the AR, AA, and NI material conditions (Table 2) were again consistent with the differences in microstructure between the specimens. For AA specimens, the lower r UTS compared to AR specimens is attributed to the precipitates in this material being coarser and more sparsely distributed (Fig. 1) and therefore offering less resistance to dislocation motion. However, given that there was virtually no difference in E or r YS between these two conditions and the difference in r UTS was only 21 MPa, it appears that the effect of the aging treatment on room-temperature tensile properties was minor. Consistent with this, Lai and Kimball [28] found that aging Alloy 800H in air at 760°C for 1000 h produced no appreciable change in r YS , and Jordan et al. [29] reported only a slight increase in r YS and a slight decrease in ductility following aging in air at 750°C for 10000 h. The latter attributed the changes to secondary carbide precipitation during aging. Although the aging times used in both these works and the present study were well short of typical service times for Alloy 800H components, these findings indicate that the fundamental microstructural changes expected to occur during service at high temperatures are unlikely to significantly affect the roomtemperature tensile performance of this alloy.
The difference in microstructure between AR and NI specimens was much greater (see Figs. 1a, c, e and 2), and this was reflected in their respective properties ( Table 2). The higher E and r YS values for NI specimens can be attributed to the effects of both increased matrix nitrogen concentration and precipitation. Increasing matrix nitrogen concentration is known to increase E in austenitic steels [30][31][32], and it also has a significant solid solution strengthening effect [30] without being as detrimental to ductility as other alloying elements such as carbon [33,34]. Highmodulus, high-stability precipitate phases such as TiC and NbC are known to increase E in austenitic steels [31], so it is feasible that the Ti(C,N) and AlN precipitated in the NI specimens also contributed to the higher values obtained for these specimens. It is also expected that precipitation of nitrides and carbonitrides would have a further strengthening effect, but several authors have observed either a decrease or very little increase in r YS and r UTS with increasing nitride phase fraction [33,35,36]. All of these studies focused on aging of high-nitrogen steels with no external nitrogen source (i.e., overall alloy composition remained the same), and all came to the conclusion that the lack of strengthening was due to depletion of nitrogen from the matrix. This implies that nitride precipitation will only serve to increase strength if the matrix nitrogen concentration is maintained.
The low r UTS of the NI specimens initially appeared to contradict the increase in r YS . However, accounting for the sudden reduction in specimen cross-sectional area due to fracture of the grain boundary Cr 2 N precipitates (Fig. 5) after yield gave r UTS values that were more consistent with the r YS results. Regardless, it is clear from the fracture of Cr 2 N precipitates and complete failure of test specimens at low strains (Fig. 4c) that the presence of both Cr 2 N and AlN severely limited the overall ductility of the NI test specimens. This implies that if in-service nitrogen uptake in Alloy 800H components reaches the point where AlN formation can occur, then their capacity for plastic deformation could be significantly impacted.

Creep properties
To understand the influence of microstructure on creep behavior, it is important to understand the deformation mechanism acting. There are four main mechanisms commonly used to characterize creep behavior of metals and alloys at temperatures above approximately 40% of the melting point: low-temperature power-law creep, where transport of matter through dislocation cores is rate-controlling; hightemperature power-law creep (dislocation climb); Coble creep (diffusion of vacancies along grain boundaries); and Nabarro-Herring creep (diffusion of vacancies through the lattice). Traditional expressions for minimum strain rate under each of these mechanisms contain the power-law relationship _ min / r n , where n is known as the stress exponent [37,38]. n = 1 for diffusion-controlled creep and n C 3 is typical of power-law creep, which lends itself to a simple method for identifying the dominant creep mechanism acting. Assuming a simplified model _ min ¼ Ar n , where A is a constant that captures the influence of temperature, grain size, and material properties, then for a given temperature and grain size a plot of log _ min vs. log r is expected to show a linear trend where n is the slope. The value of n indicates the dominant creep mechanism, and a change in slope signals a transition from one creep mechanism to another. Figure 10 shows AR specimen data from this work plotted with literature data for Alloy 800H with grain size * 100 lm creep tested at 1000°C. The raw data suggests a change in slope around 10-12 MPa applied stress. Fitting the simplified model given above yields n-values of 1.6 for r \ * 10 MPa and 4.4 for r [ * 12 MPa, indicating diffusion-controlled and power-law mechanisms, respectively. There was no change in the n-value for r \ * 10 MPa (to 1 dp) regardless of whether the data from this work was included in the fitting or not. Note that data in the range 10-12 MPa was used for both fits due to the lack of a distinct transition point. The AR specimen data lies on the line where n = 1.6, meaning that diffusion-controlled creep was likely the dominant mechanism for these tests. It is not clear whether Coble (boundary diffusion) or Nabarro-Herring (lattice diffusion) creep was rate-controlling, but the latter tends to dominate at temperatures closer to the melting point. The Coble mechanism is more likely here, as the test temperature was \ 80% of the melting temperature of 800H [4].
The test conditions in this work were originally selected using Beardsley et al.'s [18] CDMM for Alloy 800H and its underlying _ min model. This model uses a summation of the traditional power-law models for the four main creep mechanisms to account for the fact that these mechanisms do not necessarily act exclusively, particularly around the transition from one to another, and is based on a large experimental data set. For the test conditions and material grain size used in this work, Beardsley et al.'s model predicts high-temperature power-law creep to be the dominant mechanism for AR specimens. It also predicts this mechanism to dominate at test conditions of 5 MPa and 1000°C, as studied by Japan's National Research Institute for Metals (NRIM) [39] (data points shown in Fig. 10). Both of these predictions disagree with the analysis illustrated in Fig. 10. These discrepancies are most likely due to the fact that Beardsley et al.'s model was based on optimization of a large data set, where most of the experimental data lay within the low-and high-temperature power-law creep regimes and did not cover the conditions used in this work. Beardsley et al.'s analysis of the model indicated that extrapolability was generally poor outside of regions where creep test data had been gathered, and they concluded that more data was required in the Coble creep regime to validate the contribution of this mechanism.
AA specimens generally demonstrated a higher _ min than AR specimens, although the difference between the two material conditions was typically less than one order of magnitude (Fig. 7). Consistent with this, Witzke and Stephens [42] found that aging Alloy 800H specimens in argon at 760°C for 3500 h prior to creep testing at 760-815°C under constant stresses in the range 41-83 MPa resulted in an increase in _ min of around one order of magnitude or less. The scope of the present work did not extend to post-test TEM analysis of creep specimens, so a microstructural explanation of this result is not provided here. Regardless, assuming diffusional creep to be the dominant mechanism it is suggested that the higher creep rates in AA material may be primarily

This work NRIM [39]
McAllister [40] Beardsley [18] Guttmann [41] Figure 10 Comparison of AR specimen minimum strain rates to values from the literature, see Refs. [18,[39][40][41]. All tests undertaken at 1000°C in air. Dashed lines are trendlines of the form _ min ¼ Ar n fitted to the data shown. n & 1 is indicative of a diffusional creep mechanism, while n C 3 is typically considered to represent power-law creep.
due to the difference in grain boundary precipitates between AR and AA material.
Grain boundary precipitates are known to influence diffusional creep rates [43][44][45][46][47][48]. Experiments have indicated that diffusional creep rates are lower and creep activation energies are higher in materials containing grain boundary precipitates compared to the pure base metal [43,46]. Grain boundary precipitates have also been associated with a threshold stress that must be overcome for appreciable creep deformation to occur [43,[45][46][47][48][49][50]. A similar threshold stress has been observed in particle-strengthened alloys undergoing power-law creep (dislocation motion is rate-controlling) [38], which may imply that precipitate characteristics that provide improved creep strength under power-law conditions (e.g., small size and close spacing, thermal stability) may also be beneficial for diffusional creep. Several analytical treatments of diffusional creep even incorporate the concept of grain boundary dislocations that behave in much the same way as lattice dislocations but are confined to the plane of the boundary [46][47][48], and experimental observations of such boundary dislocations have been reported [46,47].
Regardless, experiments indicate that for a given volume fraction of precipitates, small grain boundary precipitates reduce diffusional creep rates more than large precipitates [43], while large precipitates appear to improve creep ductility by impeding coalescence of voids during tertiary creep [51]. Studies on aluminum alloys and austenitic stainless steels indicate that increasing the area fraction of grain boundary occupied by precipitates will also reduce creep rates [52,53]. Some of the analytical treatments mentioned previously give the threshold stress for diffusional creep as a function of grain boundary precipitate spacing (analogous to the Orowan stress for lattice dislocations), which has proved reasonably successful in prediction of experimental creep rates [46,48].
Considering the present work, these findings suggest that the (initially) lower area fraction of grain boundary M 23 C 6 precipitates/greater precipitate spacing and larger precipitate size in AA material may have been responsible for the higher _ min compared to AR material. Note that while the grainboundary M 23 C 6 in AR specimens did coarsen during testing, t min for all AR specimens was less than the 800 h aging time used for heat treatment of AA specimens. Therefore, the area fraction of grain boundary precipitates in AR material at t min is still expected to have been greater than in AA material.
All NI specimens demonstrated an _ min at least half an order of magnitude lower than that of the AR specimens (Fig. 7), which is attributed primarily to the presence of the nitride precipitates in the microstructure. This is on the basis that oxide-dispersion-strengthened (ODS) Ni-base superalloys, which are considered analogous to the system studied in this work, show significantly reduced diffusional creep rates compared to the oxide-free alloys [38,46]. Apparent threshold stresses on the order of 10 MPa have been determined for ODS Ni-based alloys tested at around 1100°C [46]. This threshold stress is comparable to the applied stress used in this work (nominally 8 MPa), so it is feasible that the applied stress in this work was close to a value where creep deformation becomes negligible. Based on the previous discussion, it is also possible that the increase in area fraction of grain boundaries occupied by precipitates due to the apparent migration of Cr 2 N precipitates during creep testing (Fig. 9) contributed to the low measured creep rates. In addition, increasing matrix nitrogen concentration has been found to have a considerable strengthening effect in austenitic stainless steels under power-law creep conditions [32,54,55]. No references for the effects of nitrogen on diffusional creep have been identified in the literature, but under the assumption that diffusional creep is influenced by similar factors to powerlaw creep, it is feasible that increased matrix nitrogen concentration will also improve creep strength.
In the AR09-06 and AA09-04 specimens, the decreases in creep strain rates shortly before test termination (Fig. 6a, b) are attributed to nitrogen uptake and nitride formation in these specimens (Fig. 8). Nitride precipitation around large creep cracks during creep testing of Alloy 800H in air has been observed by other authors, see [41,56]. Welker et al. [57] used compact test specimens to determine that the critical crack opening rate for nitride formation during testing in air at 1000°C is between 1 9 10 -6 s -1 . The maximum strain rate recorded for the AR09-06 specimen was approximately 1 9 10 -7 s -1 , indicating that the rate of crack opening was within the critical range identified by Welker et al.
Welker et al. [57] also observed that the rate of nitride penetration around the creep crack was comparable to that observed in unstressed specimens during aging in a 95% N 2 ? 5% H 2 atmosphere at 1000°C. We previously characterized the kinetics of AlN and Cr 2 N penetration in Alloy 800H under these same conditions [13]. Taking the crack surface as x = 0, AlN penetration around 6 cracks in the AR09-06 specimen was measured and the average was found to be 683 lm. Using the kinetic model for AlN penetration developed in [13], this indicates that AlN formation began around 84 h prior to test termination (83-89 h at a 95% confidence level). From the strain vs. time data shown in Fig. 6a, it was estimated that the creep strain rate began to decrease around 100 h before test termination, which agrees reasonably well with this prediction.
The AlN precipitates in the AA09-04 specimen (Fig. 8b) were not associated with creep cracks, and instead appeared to have formed as a result of nitrogen ingress from the free surface. This indicates that the rate of deformation at the free surface was high enough to damage the surface oxide scale to the point where some level of nitrogen uptake was possible. Similar AlN precipitation behavior has been observed by Erneman et al. [6], who reported through-thickness AlN formation in 5 mm-diameter Alloy 800HT specimens creep tested in air at 1000°C and 16 MPa for 347 h. In this work, penetration depth in the AA09-04 specimen was approximately 400 lm, and from Fig. 6b it is estimated that nitride formation may have begun around 150 h prior to test termination. This indicates a slower rate of penetration than around the creep cracks in AR09-06, meaning that the models developed in [13] will not correctly predict penetration under these circumstances. Regardless, these results show that nitride formation can occur in Alloy 800H in an air atmosphere under creep conditions, and that cracking is not a pre-requisite.

Implications for service
The observed differences in mechanical performance between as-received, aged, and nitrided Alloy 800H material may have serious implications for both design lifetime calculation and more importantly, inservice fitness-for-service assessment. In this work, we have shown that Alloy 800H can suffer a decrease in creep strength (increase in _ min ) due to aging. Rupture times were not obtained from the tests conducted, meaning LMP values could not be calculated and compared with those predicted by the API 530 model, but it is expected that there is an inverse relationship between _ min and t r according to the Monkman-Grant relationship [58]. On this basis, an increase in _ min with aging implies that the API 530 [2] property model will overestimate lifetime from the outset, and that the design lifetime of 100000 h cannot be reached at design conditions.
A decrease in creep strength with aging also implies that API 579-1 fitness-for-service assessments conducted toward end-of-life will overestimate remaining life due to creep strength being overestimated. In a worst-case scenario, this could lead to an unexpected failure. However, given that the difference in minimum creep rate between AR and AA material was less than an order of magnitude, it is possible that this change may be captured by the uncertainties in the API 579-1 creep model for 800H.
To evaluate whether the change in creep strength with aging observed in this work was significant enough to render the industry-standard API 579-1 Omega model for Alloy 800H invalid, this model was fitted to the creep test data for AR09-06 and AA09-04. The fitted parameters and expected bounds are given in Table 3. Both specimens lay within the expected bounds. AR09-06 was close to the upper bound for strain rate (corresponding to high creep strength) and displayed good ductility (D cd = -0.3 corresponds to ductile behavior and ? 0.3 to brittle behavior), while AA09-04 was nearer the lower bound for strength and close to average for ductility. Regardless, these results indicate that the difference in creep performance between the AR and AA specimens is comparable to normal scatter in the creep properties of Alloy 800H. While there was clearly a decrease in creep strength with aging, the implication is that this decrease is not significant enough to require a more complex material property model for use in remaining life assessments that accounts for this aging effect. However, we recommend that lower bound properties, particularly for D sr , should be used as the default for Alloy 800H components that have seen particularly high temperatures and/or a long time in service.
The Omega model was not fitted to NI specimen data as a clear tertiary stage region of the creep curve is required, which was not obtained for these specimens. Despite this, it is highly probable that all NI specimens will have fallen well outside the expected bounds for D sr , and likely also for D cd . For example, all NI specimens demonstrated greater creep strength than AR09-06, which fell close to the API 579-1 upper bound for strength. This implies that the API 579-1 model will not satisfactorily capture the creep behavior of nitrided 800H and should not be used in the absence of experimental creep data for remaining life assessment of material that is known or suspected to have suffered nitriding. Further work beyond the scope of this study is required to develop a suitable model for this situation and/or to determine a threshold for removal of a component from service.
The severely reduced tensile ductility of nitrided 800H at room temperature also raises concerns for fitness-for-service. Pigtails provide a connection between large-diameter reformer tubes and collection manifolds, and are typically designed to accommodate thermal expansion of these components during start-up and shutdown of the reformer [9]. This requires them to have good ductility over a wide range of temperatures. Reduced ductility due to nitridation therefore puts them at greater risk of overload failure during a temperature cycle. The risk of stress concentrations developing into cracks that lead to rupture is also likely to be greater, as low ductility is often associated with low fracture toughness. Characterization of fracture toughness of nitrided material and how fracture toughness and ductility change with temperature is again beyond the scope of this study, but it is likely to be a topic worthy of further investigation.

Summary and conclusions
In this work, we characterize the microstructures of as-received, aged, and nitrided Alloy 800H, an austenitic stainless steel commonly used in high-temperature industrial applications, and show that the microstructural changes induced by aging and nitriding are associated with changes in both roomtemperature deformation behavior and high-temperature creep behavior. To the best of our knowledge, this is the first time that a study specifically focusing on the effects of nitriding on the tensile and creep behavior of Alloy 800H has been presented in the published literature.
Using optical, SEM, and TEM techniques, we compared the microstructures of as-received (AR) samples of annealed Alloy 800H tube material, aged (AA) samples that had been subject to heat treatment in 99.99% argon for 800 h at 1000°C and 1 atm. pressure, and nitrided (NI) samples that had been exposed to a 95% N 2 ? 5% H 2 atmosphere under the same conditions for the same period of time. Roomtemperature hardness and tensile tests, and hightemperature creep tests at 1000°C and 8.3 MPa in air were undertaken on specimens in each material condition to evaluate the impact of aging and nitriding on mechanical properties.
Based on the findings of this work, the following conclusions are drawn: • Aging of annealed Alloy 800H tube material in an inert atmosphere at temperatures close to typical pigtail service conditions can cause some coarsening of primary precipitates, but otherwise does not cause significant changes in the microstructure.
• Consistent with this, the effect of aging under these conditions on room temperature hardness and tensile properties is limited to minor decreases in hardness and ultimate tensile strength. • Creep strain rates in aged material are higher than for as-received material, and it is believed that this is primarily due to differences in grain boundary precipitate size and morphology. • Nitriding of the same original Alloy 800H material leads to formation of the nitride phases AlN, Cr 2 N, and CrN, and secondary precipitation of M 23 C 6 and Ti(C,N). • Nitride precipitation and higher levels of dissolved nitrogen in the alloy matrix are attributed with increasing room temperature hardness, elastic modulus, and yield strength, and severely decreasing ductility in comparison to as-received material. • Nitrided Alloy 800H material demonstrates lower creep strain rates (higher creep strength) than asreceived or aged material, and it is believed that the nitride precipitates are primarily responsible for this behavior in a manner similar to the strengthening effects of oxide particles in oxidedispersion-strengthened alloys. • Nitride precipitation can occur in as-received and aged material subject to creep in an air atmosphere, and formation of these precipitates is associated with a decrease in creep strain rate. • The measured creep behavior of both as-received and aged Alloy 800H material falls within the bounds predicted by the MPC Project Omega creep model for Alloy 800H provided in the API 579-1 Fitness-For-Service standard, which is widely used in industry for assessment of fitness-for-service and remaining life. This indicates that the API 579-1 model is appropriate for both new material and material that has been in service for some time, although it is recommended that lower-bound properties be used for ex-service material. • Insufficient data were collected from nitrided material to allow fitting of the Omega model for this material condition, but it is expected that the creep behavior of nitrided Alloy 800H will fall well outside the API 579-1 bounds for this alloy. This implies that the API 579-1 model cannot reliably be used to assess material that is known or suspected to have suffered nitriding without supplementary experimental data to verify the creep properties of the affected material.
otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.