Analysis of the Thermomechanical Fatigue Behavior of Fully Ferritic High Chromium Steel Crofer ® 22 H with Cyclic Indentation Testing

: The 22 wt.% Cr, fully ferritic stainless steel Crofer ® 22 H has higher thermomechanical fatigue (TMF)- lifetime compared to advanced ferritic-martensitic P91, which is assumed to be caused by di ﬀ erent damage tolerance, leading to di ﬀ erences in crack propagation and failure mechanisms. To analyze this, instrumented cyclic indentation tests (CITs) were used because the material’s cyclic hardening potential—which strongly correlates with damage tolerance, can be determined by analyzing the deformation behavior in CITs. In the presented work, CITs were performed for both materials at specimens loaded for di ﬀ erent numbers of TMF-cycles. These investigations show higher damage tolerance for Crofer ® 22 H and demonstrate changes in damage tolerance during TMF-loading for both materials, which correlates with the cyclic deformation behavior observed in TMF-tests. Furthermore, the results obtained at Crofer ® 22 H indicate an increase of damage tolerance in the second half of TMF-lifetime, which cannot be observed for P91. Moreover, CITs were performed at Crofer ® 22 H in the vicinity of a fatigue crack, enabling to locally analyze the damage tolerance. These CITs show di ﬀ erences between crack edges and the crack tip. Conclusively, the presented results demonstrate that CITs can be utilized to analyze TMF-induced changes in damage tolerance. leads to the observed changes in cyclic hardening potential, as well as hardness. Future research aims to investigate the interrelation between dislocation density, Laves phase particles, and mechanical properties.


Introduction
Because of the current dramatic changes in electricity generation and supply, conventional thermal power plants are operated in a more fluctuating way, resulting in a combination of creep and cyclic loading of the hot section components. As shown by Holdsworth et al. [1], this leads to accelerated crack propagation, and hence, reduces lifetime compared to pure creep loadings. For combined loading conditions, preliminary work at materials for thermal power plant applications [2] has shown, that 22 wt.% Cr, fully ferritic stainless steel Crofer ® 22 H exhibits a higher lifetime compared to advanced ferritic-martensitic (AFM) steel P91.
While strengthening of P91 is mainly achieved by M 23 C 6 (M: Mo, Cr)-and MX (M: V, Nb; X: N)-particles, as well as solid solution and sub-grain hardening [3], the fully ferritic Crofer ® 22 H, can be strengthened by a combination of solid solution hardening and intermetallic Laves phase (Fe,Cr,Si) 2 (Nb,W) precipitates [4], whereby Crofer ® 22 H has a limited solubility of C and N in the ferritic matrix. For Crofer ® 22 H, the decisive factor for hardening in (thermomechanical) fatigue

Materials
The advanced ferritic-martensitic (AFM) steel P91, belonging to the 9Cr-1Mo family of steels, is widely applied in current thermal power plants [23,24]. P91 has a tempered martensite structure after austenitization, air cooling, and tempering. As shown in Figure 1, the tempered martensite is characterized by prior austenite grain boundaries (PAGBs), as well as packet and lath boundaries [3]. The stability is achieved by precipitation (predominantly Cr and Mo-rich M 23 C 6 , M 7 C 3 , M 3 C, M 2 X, Cr 2 N), solid solution, and sub-grain hardening [3]. M 23 C 6 and MX precipitate along PAGBs lath boundaries, packet boundaries (see Figure 1), as well as inside intra-lath regions [3]. However, long-term stability during creep stress is deteriorated by the formation of the intermetallic Laves phase and Z-phase [3].

Materials
The advanced ferritic-martensitic (AFM) steel P91, belonging to the 9Cr-1Mo family of steels, is widely applied in current thermal power plants [23,24]. P91 has a tempered martensite structure after austenitization, air cooling, and tempering. As shown in Figure 1, the tempered martensite is characterized by prior austenite grain boundaries (PAGBs), as well as packet and lath boundaries [3]. The stability is achieved by precipitation (predominantly Cr and Mo-rich M23C6, M7C3, M3C, M2X, Cr2N), solid solution, and sub-grain hardening [3]. M23C6 and MX precipitate along PAGBs lath boundaries, packet boundaries (see Figure 1), as well as inside intra-lath regions [3]. However, longterm stability during creep stress is deteriorated by the formation of the intermetallic Laves phase and Z-phase [3]. The characteristic microstructure of tempered ferritic-martensitic steel P91 [5]. PAGB, prior austenite grain boundaries.
The P91 investigated in the present study was extracted from a pipe section with an outer diameter of 160 mm and a wall thickness of 20 mm. Austenitization was carried out at 1050 °C for 0.5 h, followed by air cooling. Afterwards, tempering at 750 °C for 1 h and subsequent air cooling were performed.
The additionally investigated Crofer ® 22 H is a 22 wt.% Cr, fully ferritic stainless steel and was developed by a collaboration between Forschungszentrum Juelich GmbH, Germany, and VDM Metals GmbH, Germany [25]. It serves as a low-cost, metallic interconnector material for the lightweight design of solid oxide fuel cell stacks [26]. Figure 2 depicts a typical microstructure of recrystallized (1075 °C/22 min/air cooling), and subsequently, precipitation annealed (650 °C/2 h/water quenching) Crofer ® 22 H. Fine Laves phase particles are uniformly dispersed [27,28] in the matrix and additionally, the high angle grain boundaries (HAGBs) are covered by Laves phase. Furthermore, particle-free zones (PFZs) are formed along the HAGBs. The strength of the material is based on a combination of solid solution hardening and intermetallic (Fe,Cr,Si)2(Nb,W) Laves phase particle precipitation [4]. Moreover, thermomechanically triggered precipitation of the Laves phase particles [29,30] has a significant impact on TMF-behavior [5], and fatigue performance [6] of such Laves phase strengthened, ferritic steels. In addition to that, an excellent creep behavior in the temperature range of 600-650 °C has been demonstrated by Reference [4] for this type of steels. The P91 investigated in the present study was extracted from a pipe section with an outer diameter of 160 mm and a wall thickness of 20 mm. Austenitization was carried out at 1050 • C for 0.5 h, followed by air cooling. Afterwards, tempering at 750 • C for 1 h and subsequent air cooling were performed.
The additionally investigated Crofer ® 22 H is a 22 wt.% Cr, fully ferritic stainless steel and was developed by a collaboration between Forschungszentrum Juelich GmbH, Germany, and VDM Metals GmbH, Germany [25]. It serves as a low-cost, metallic interconnector material for the light-weight design of solid oxide fuel cell stacks [26]. Figure 2 depicts a typical microstructure of recrystallized (1075 • C/22 min/air cooling), and subsequently, precipitation annealed (650 • C/2 h/water quenching) Crofer ® 22 H. Fine Laves phase particles are uniformly dispersed [27,28] in the matrix and additionally, the high angle grain boundaries (HAGBs) are covered by Laves phase. Furthermore, particle-free zones (PFZs) are formed along the HAGBs. The strength of the material is based on a combination of solid solution hardening and intermetallic (Fe,Cr,Si) 2 (Nb,W) Laves phase particle precipitation [4]. Moreover, thermomechanically triggered precipitation of the Laves phase particles [29,30] has a significant impact on TMF-behavior [5], and fatigue performance [6] of such Laves phase strengthened, ferritic steels. In addition to that, an excellent creep behavior in the temperature range of 600-650 • C has been demonstrated by Reference [4] for this type of steels. In the presented work, recrystallized (1050 °C/5+ minutes, depending on section thickness/rapid air-cooling) Crofer ® 22 H from VDM Metals, Germany was investigated.
The chemical compositions of Crofer ® 22 H were determined with inductively coupled plasma optical emission spectrometry and are given in Table 1, which also contains the chemical composition from the manufacturer's datasheet of P91. In Table 2, the typical mechanical properties at an ambient temperature of the as-delivered state are shown, respectively.

Methods
Total strain-controlled out-of-phase (oop) thermomechanical fatigue (TMF-) tests were performed in accordance with the European Code-of-Practice [33]. Therefore, cylindrical specimens with a gauge length of 15 mm and a diameter of 7 mm, as well as servo-hydraulic fatigue testing systems, equipped with inductive specimen heating systems, were used. For temperature control, a type R ribbon thermocouple was applied in the center of the gauge length. The strain was directly measured using high-temperature extensometers. All experiments were conducted in a temperature range from 250 to 650 °C with heating, as well as the cooling rate of dT/dt = 10 Ks −1 . To reduce creep deformation to a minimum, no holding time was applied. To achieve the specified cooling rate, the specimens were directly cooled with compressed air after reaching the maximum temperature. Moreover, the resulting thermal strain εth was suppressed by applying a mechanical strain of εmech = −εth in oop TMF cycles. Since the TMF cycle started at a minimum temperature of 250 °C, the strain In the presented work, recrystallized (1050 • C/5+ minutes, depending on section thickness/rapid air-cooling) Crofer ® 22 H from VDM Metals, Germany was investigated.
The chemical compositions of Crofer ® 22 H were determined with inductively coupled plasma optical emission spectrometry and are given in Table 1, which also contains the chemical composition from the manufacturer's datasheet of P91. In Table 2, the typical mechanical properties at an ambient temperature of the as-delivered state are shown, respectively.

Methods
Total strain-controlled out-of-phase (oop) thermomechanical fatigue (TMF-) tests were performed in accordance with the European Code-of-Practice [33]. Therefore, cylindrical specimens with a gauge length of 15 mm and a diameter of 7 mm, as well as servo-hydraulic fatigue testing systems, equipped with inductive specimen heating systems, were used. For temperature control, a type R ribbon thermocouple was applied in the center of the gauge length. The strain was directly measured using high-temperature extensometers. All experiments were conducted in a temperature range from 250 to 650 • C with heating, as well as the cooling rate of dT/dt = 10 Ks −1 . To reduce creep deformation to a minimum, no holding time was applied. To achieve the specified cooling rate, the specimens were directly cooled with compressed air after reaching the maximum temperature. Moreover, the resulting thermal strain ε th was suppressed by applying a mechanical strain of ε mech = −ε th in oop TMF cycles. Since the TMF cycle started at a minimum temperature of 250 • C, the strain ratio of mechanical strain was R ε, mech = −∞. The differently pronounced thermal expansion of P91 and Crofer ® 22 H, which is mainly caused by substantially different Cr content, leads to deviations in ε th , and thus, in ε mech . While P91 was loaded with ε mech = 0.54%, Crofer ® 22 H was charged with ε mech = 0.509%. Consequently, one may expect a higher stress range for P91, but in TMF tests, a higher stress level was observed for Crofer ® 22 H (see Figure 3), caused by pronounced Laves phase strengthening of Crofer ® 22 H (see Reference [5]). Moreover, both materials show a decrease in stress range during TMF-loading, which is more pronounced for Crofer22 H. The results of TMF-experiments are discussed in detail in Reference [5].
Appl. Sci. 2020, 10, x FOR PEER REVIEW 5 of 13 ratio of mechanical strain was Rε, mech = −∞. The differently pronounced thermal expansion of P91 and Crofer ® 22 H, which is mainly caused by substantially different Cr content, leads to deviations in εth, and thus, in εmech. While P91 was loaded with εmech = 0.54 %, Crofer ® 22 H was charged with εmech = 0.509 %. Consequently, one may expect a higher stress range for P91, but in TMF tests, a higher stress level was observed for Crofer ® 22 H (see Figure 3), caused by pronounced Laves phase strengthening of Crofer ® 22 H (see Reference [5]). Moreover, both materials show a decrease in stress range during TMF-loading, which is more pronounced for Crofer22 H. The results of TMF-experiments are discussed in detail in Reference [5]. Despite slightly higher maximum stress, Crofer ® 22 H exhibits in Reference [5] higher TMFlifetime in relation to P91 (see Figure 3), which might be explained by differently pronounced damage tolerance. To analyze the changes in cyclic hardening potential, which correlates with damage tolerance [17,18], during TMF-loading, instrumented cyclic indentation tests (CITs) were conducted at the initial state (IS), as well as after defined numbers of TMF-cycles NTMF. Therefore, the CITs were performed in the volume of the gauge length of the respective specimen with a maximum indentation force of Fmax = 1000 mN and in total ten indentation cycles with a frequency of f = 1/12 Hz. Because this requires a destructive preparation of the specimens' gauge length, different specimens were loaded for various NTMF. For CITs, a Fischerscope H100C (Helmut Fischer GmbH, Germany), which enables continuous detection of the indentation depth h and indentation force F, was used.
Additionally, CITs were performed at the crack edges, as well as in front of the crack tip, and thus, in the plastic zone, which can be estimated according to Reference [34]. These measurements were conducted at a fatigue crack, observed at a specimen of Crofer ® 22 H loaded for 4342 TMF-cycles (see Figure 4a). In addition to Fmax = 1000 mN, a maximum indentation force of 100 mN was used to get a higher resolution of the material volume around the fatigue crack (see Reference [16]). Despite slightly higher maximum stress, Crofer ® 22 H exhibits in Reference [5] higher TMF-lifetime in relation to P91 (see Figure 3), which might be explained by differently pronounced damage tolerance. To analyze the changes in cyclic hardening potential, which correlates with damage tolerance [17,18], during TMF-loading, instrumented cyclic indentation tests (CITs) were conducted at the initial state (IS), as well as after defined numbers of TMF-cycles N TMF . Therefore, the CITs were performed in the volume of the gauge length of the respective specimen with a maximum indentation force of F max = 1000 mN and in total ten indentation cycles with a frequency of f = 1/12 Hz. Because this requires a destructive preparation of the specimens' gauge length, different specimens were loaded for various N TMF . For CITs, a Fischerscope H100C (Helmut Fischer GmbH, Germany), which enables continuous detection of the indentation depth h and indentation force F, was used.
Additionally, CITs were performed at the crack edges, as well as in front of the crack tip, and thus, in the plastic zone, which can be estimated according to Reference [34]. These measurements were conducted at a fatigue crack, observed at a specimen of Crofer ® 22 H loaded for 4342 TMF-cycles (see Figure 4a). In addition to F max = 1000 mN, a maximum indentation force of 100 mN was used to get a higher resolution of the material volume around the fatigue crack (see Reference [16]). The CITs were evaluated using the PhyBaLCHT procedure presented in detail in References [13,15,16]. During indentation, the indentation force F and indentation depth h are measured continuously, revealing an F-h hysteresis from the 2nd cycle on (see Figure 4b). The half-width of the F-h hysteresis at mean loading is considered as a characteristic parameter to describe the deformation behavior and is called plastic indentation depth amplitude ha,p. The ha,p-N relation plotted schematically in Figure 4c, shows a slope change from the 5th cycle, which indicates saturation of macroplastic deformation, and hence, the deformation behavior in CITs is dominated by microplasticity in this regime. Consequently, the regime after the 5th cycle, which can further be described by the power-law function ha,p II (see Equation (1)), is analyzed to characterize the material's cyclic properties [15].
The slope of ha,p II indicates the amount of cyclic hardening during CIT and is described by the exponent eII, called cyclic hardening exponentCHT [15]. A steeper slope results in a higher |eII|, which implies a larger amount of cyclic hardening during indentation testing. In addition to that and in correspondence to conventional microhardness measurement, the Martens hardness (HM) can be determined from the F-h relation of the 1st cycle.
To obtain statistically reliable values of ha,p-N curves, and HM, for each condition, mean values were determined using 40 indentation points, respectively. Note that for the measurements in the relatively small plastic zone in front of the crack tip, only 20 indentation points could be realized. For investigating the mechanical properties at the crack edges, 40 indentation points were utilized with Fmax = 1000 mN, whereas a Fmax = 100 mN enabled the possibility to perform 60 CITs along the crack edge. For analysis in the fatigue crack vicinity, CITs were performed along the fatigue crack (comp. Figure 4a) and on different planes prepared by grinding and polishing the sample. In Section 4, for each condition, the mean values and 90% confidence intervals, based on the overall number of indentation points, are shown.

Evolution of Cyclic Hardening Potential and Hardness During TMF Loading
The TMF-tests demonstrate, in spite of slightly larger stress range and mean stresses (91 MPa for CroFer ® 22 H and 75 MPa for P91), a higher lifetime for CroFer ® 22 H compared to P91 (see Figure 3). While P91 exhibits a strong degradation of its initial martensitic lath structure, called The CITs were evaluated using the PhyBaL CHT procedure presented in detail in References [13,15,16]. During indentation, the indentation force F and indentation depth h are measured continuously, revealing an F-h hysteresis from the 2nd cycle on (see Figure 4b). The half-width of the F-h hysteresis at mean loading is considered as a characteristic parameter to describe the deformation behavior and is called plastic indentation depth amplitude h a,p . The h a,p -N relation plotted schematically in Figure 4c, shows a slope change from the 5th cycle, which indicates saturation of macroplastic deformation, and hence, the deformation behavior in CITs is dominated by microplasticity in this regime. Consequently, the regime after the 5th cycle, which can further be described by the power-law function h a,p II (see Equation (1)), is analyzed to characterize the material's cyclic properties [15].
The slope of h a,p II indicates the amount of cyclic hardening during CIT and is described by the exponent e II , called cyclic hardening exponent CHT [15]. A steeper slope results in a higher |e II |, which implies a larger amount of cyclic hardening during indentation testing. In addition to that and in correspondence to conventional microhardness measurement, the Martens hardness (HM) can be determined from the F-h relation of the 1st cycle.
To obtain statistically reliable values of h a,p -N curves, and HM, for each condition, mean values were determined using 40 indentation points, respectively. Note that for the measurements in the relatively small plastic zone in front of the crack tip, only 20 indentation points could be realized. For investigating the mechanical properties at the crack edges, 40 indentation points were utilized with F max = 1000 mN, whereas a F max = 100 mN enabled the possibility to perform 60 CITs along the crack edge. For analysis in the fatigue crack vicinity, CITs were performed along the fatigue crack (comp. Figure 4a) and on different planes prepared by grinding and polishing the sample. In Section 4, for each condition, the mean values and 90% confidence intervals, based on the overall number of indentation points, are shown.

Evolution of Cyclic Hardening Potential and Hardness During TMF Loading
The TMF-tests demonstrate, in spite of slightly larger stress range and mean stresses (91 MPa for CroFer ® 22 H and 75 MPa for P91), a higher lifetime for CroFer ® 22 H compared to P91 (see Figure 3). While P91 exhibits a strong degradation of its initial martensitic lath structure, called "polygonization", the ferritic CroFer ® 22 H shows an increase of the Laves phase particle fraction, caused by TMF-loading and leading to a higher stress range and maximum stress in relation to P91 (see Figure 3 and [5]). These microstructural phenomena described in Reference [5] are dependent on the number of TMF-cycles N TMF and influence the material's damage tolerance. To determine the evolution of damage tolerance during TMF-loading, CITs were performed at different N TMF .
Although the TMF-tests were performed at variable, elevated temperatures, and thus, at different conditions compared to the CITs performed at ambient temperature, the CITs enable the determination of the evolution of the material's mechanical properties, i.e., hardness and cyclic hardening potential, resulting from TMF-loading. Note that cyclic hardening potential, represented by e II , strongly correlates with the material's damage tolerance [17,18].
Comparing the materials in their initial states (IS), Crofer ® 22 H has lower hardness, but significantly higher cyclic hardening potential, compared to P91 (see Figure 5)-which correlates to stronger damage tolerance indicated in Reference [5]. Moreover, Crofer ® 22 H shows lower h a,p values in relation to P91, due to extensively more pronounced cyclic hardening during cyclic indentation (see Figure 6).
Appl. Sci. 2020, 10, x FOR PEER REVIEW 7 of 13 "polygonization", the ferritic CroFer ® 22 H shows an increase of the Laves phase particle fraction, caused by TMF-loading and leading to a higher stress range and maximum stress in relation to P91 (see Figure 3 and [5]). These microstructural phenomena described in Reference [5] are dependent on the number of TMF-cycles NTMF and influence the material's damage tolerance. To determine the evolution of damage tolerance during TMF-loading, CITs were performed at different NTMF.
Although the TMF-tests were performed at variable, elevated temperatures, and thus, at different conditions compared to the CITs performed at ambient temperature, the CITs enable the determination of the evolution of the material's mechanical properties, i.e., hardness and cyclic hardening potential, resulting from TMF-loading. Note that cyclic hardening potential, represented by eII, strongly correlates with the material's damage tolerance [17,18].
Comparing the materials in their initial states (IS), Crofer ® 22 H has lower hardness, but significantly higher cyclic hardening potential, compared to P91 (see Figure 5)-which correlates to stronger damage tolerance indicated in Reference [5]. Moreover, Crofer ® 22 H shows lower ha,p values in relation to P91, due to extensively more pronounced cyclic hardening during cyclic indentation (see Figure 6).  For P91 a pronounced decrease of |e II | can be observed during the first 250 TMF-cycles, which is accompanied by a hardness decrease, correlating with the cyclic softening observed in TMF-experiments (comp. Figure 4, Figure 5 and [5]). However, from the 250th TMF-cycle on, P91 exhibits a stabilization of both hardness and |e II |, which also correlates to the cyclic deformation curves shown in Figure 3, as well as the h a,p -N curves in Figure 6. The degradation of martensite laths discussed in Reference [5] is most pronounced in the regime after the 250th cycle, and consequently, seems to have no significant impact on the mechanical properties determined in CITs, and thus, cyclic hardening potential. However, crack propagation can be influenced by this "polygonization", which cannot be detected in CITs and has, hence, to be considered by evaluating the TMF-behavior. For P91 a pronounced decrease of |eII| can be observed during the first 250 TMF-cycles, which is accompanied by a hardness decrease, correlating with the cyclic softening observed in TMFexperiments (comp. Figure 4, Figure 5 and [5]). However, from the 250th TMF-cycle on, P91 exhibits a stabilization of both hardness and |eII|, which also correlates to the cyclic deformation curves shown in Figure 3, as well as the ha,p-N curves in Figure 6. The degradation of martensite laths discussed in Reference [5] is most pronounced in the regime after the 250th cycle, and consequently, seems to have no significant impact on the mechanical properties determined in CITs, and thus, cyclic hardening potential. However, crack propagation can be influenced by this "polygonization", which cannot be detected in CITs and has, hence, to be considered by evaluating the TMF-behavior.
In CITs, CroFer ® 22 H exhibit a significant increase of hardness and decrease of |eII| in the first 20 TMF-cycles, which is accompanied by a pronounced cyclic hardening indicated by increasing stress range in Figure 3. With increasing NTMF, in spite of decreasing stress range (see Figure 3), the increase in Martens hardness and decrease in cyclic hardening potential continues, leading to a maximum of hardness and minimum of |eII| at NTMF = 1000. These extensive changes in the material properties are accompanied by a pronounced shift of ha,p-N curves to higher values at NTMF = 250 (see Figure 6), resulting in decreasing hardening potential in this condition. However, after 2000 TMF cycles, Crofer ® 22 H exhibits a decrease in hardness, but an increasing |eII|. Consequently, the damage tolerance of Crofer ® 22 H seems to be improved within the 2 nd half of TMF-lifetime, i.e., NTMF > 2000. The increase of cyclic hardening potential, and hence, damage tolerance, of Crofer ® 22 H in this state corresponds to the prolonged damage phase in TMF-test reported in Reference [5].
To explain the results observed in CITs at Crofer ® 22 H, different microstructural phenomena have to be considered. Because of the massive plastic deformation at the beginning of the TMFexperiments, an increase of dislocation density, leading to smaller |eII| and higher hardness (see Reference [16]), is assumed. In addition to that, the results in Reference [5] demonstrate an accelerated Laves phase formation caused by TMF-loading. Note that higher dislocation density would be expected to be non-permanent in TMF-condition, but because Laves phase precipitates pin the dislocations, it is assumed to be quasi-permanent [5]. The number of the Laves phase particles significantly increases in the first 2000 cycles, and afterwards, decreases until failure, which at least roughly correlates with the hardness maximum observed in CITs, but also with the respective minimum in cyclic hardening potential (comp. Figure 5 and Reference [5]). Besides a decrease in a In CITs, CroFer ® 22 H exhibit a significant increase of hardness and decrease of |e II | in the first 20 TMF-cycles, which is accompanied by a pronounced cyclic hardening indicated by increasing stress range in Figure 3. With increasing N TMF , in spite of decreasing stress range (see Figure 3), the increase in Martens hardness and decrease in cyclic hardening potential continues, leading to a maximum of hardness and minimum of |e II | at N TMF = 1000. These extensive changes in the material properties are accompanied by a pronounced shift of h a,p -N curves to higher values at N TMF = 250 (see Figure 6), resulting in decreasing hardening potential in this condition. However, after 2000 TMF cycles, Crofer ® 22 H exhibits a decrease in hardness, but an increasing |e II |. Consequently, the damage tolerance of Crofer ® 22 H seems to be improved within the 2 nd half of TMF-lifetime, i.e., N TMF > 2000. The increase of cyclic hardening potential, and hence, damage tolerance, of Crofer ® 22 H in this state corresponds to the prolonged damage phase in TMF-test reported in Reference [5].
To explain the results observed in CITs at Crofer ® 22 H, different microstructural phenomena have to be considered. Because of the massive plastic deformation at the beginning of the TMF-experiments, an increase of dislocation density, leading to smaller |e II | and higher hardness (see Reference [16]), is assumed. In addition to that, the results in Reference [5] demonstrate an accelerated Laves phase formation caused by TMF-loading. Note that higher dislocation density would be expected to be non-permanent in TMF-condition, but because Laves phase precipitates pin the dislocations, it is assumed to be quasi-permanent [5]. The number of the Laves phase particles significantly increases in the first 2000 cycles, and afterwards, decreases until failure, which at least roughly correlates with the hardness maximum observed in CITs, but also with the respective minimum in cyclic hardening potential (comp. Figure 5 and Reference [5]). Besides a decrease in a number of the Laves phase particles, a pronounced coarsening of the Laves phase particles can be observed at N TMF > 2000, which obviously leads to an increase of |e II | and lower hardness.
In this context, the results of Reference [13] for precipitation hardened Cu alloyed steels show, that the hardness, as well as the cyclic hardening potential, depend on the number and size, as well as the coherency state of Cu precipitates. Consequently, the Laves phase evolution has to be considered to explain the results observed in CITs. It is assumed that the combination of the Laves phase particle evolution and cyclic plastic deformation in TMF-tests leads to the observed changes in cyclic hardening potential, as well as hardness. Future research aims to investigate the interrelation between dislocation density, Laves phase particles, and mechanical properties.
Comparing the different materials, Crofer ® 22 H shows more pronounced changes in hardness and |e II |, due to TMF-loading (see Figure 5), which further can be seen in the h a,p -N curves (see Figure 6). While P91 has a higher hardness and smaller |e II | in the initial state, an oop TMF-loading of approximately 1000 cycles yields higher hardness and lower cyclic hardening potential for Crofer ® 22 H. However, at N TMF > 2000, Crofer ® 22 H shows increasing cyclic hardening potential, whereas P91 exhibits a stabilized state, underlining the higher damage tolerance of Crofer ® 22 H described in Reference [5]. Furthermore, the h a,p values obtained for Crofer ® 22 H are smaller compared to the ones of P91 (see Figure 6), which is less pronounced at higher N TMF , demonstrating the more extensive microstructural changes of Crofer ® 22 H during TMF-loading.
As a summary, the results observed in CITs correspond to the cyclic deformation behavior obtained in TMF-tests. Moreover, for Crofer ® 22 H, microstructural changes correlate with changes in cyclic hardening potential and hardness obtained in CITs. Furthermore, the results of CITs underline the assumption, that Crofer ® 22 H has a significantly higher damage tolerance. Additionally, Crofer ® 22 H shows in relation to P91 more pronounced changes in damage tolerance during TMF-loading, i.e., an increasing cyclic hardening potential during TMF-test, leading to a longer lifetime.

Investigation of the Local Cyclic Hardening Potential and Hardness in the Vicinity of a Fatigue Crack
As shown in Reference [5] and the results of the CITs given above, the damage tolerance of Crofer ® 22 H is significantly higher, and TMF-induced changes are more pronounced compared to P91. Moreover, the discussion in Section 4.1 demonstrates, that the evolution of cyclic hardening potential of Crofer ® 22 H is strongly dependent on dislocation density and Laves phase particles, which relate to the amount of plastic deformation, being significantly higher at the crack tip. Therefore, CITs were performed at the crack edges, as well as the plastically deformed zone in front of the crack tip of a secondary fatigue crack observed at the Crofer ® 22 H specimen loaded for N TMF = 4342 (see Figure 4a). These locally performed CITs enable a detailed analysis of the evolution of damage tolerance during fatigue crack propagation.
The results obtained with F max = 1000 mN shown on the left-hand side of Figure 7, exhibit slightly higher hardness and a significantly reduced cyclic hardening potential in front of the crack tip in relation to the material's volume in the gauge length. This is assumed to be caused by more pronounced plastic deformation in front of the fatigue crack. In correspondence to that, the crack edges also show decreased cyclic hardening potential. Due to higher amounts of plastic deformation, Laves phase particle formation is assumed to be more pronounced at the crack tip and edges (see Reference [5]), which also influences hardness, as well as |e II | and is objective of further investigation.
To analyze the differences between the crack tip and crack edge, CITs with F max = 100 mN were conducted, leading to a higher spatial resolution. Due to the indentation size effect (see References [16,22]), the CITs with F max = 100 mN yield a higher hardness compared to F max = 1000 mN (see Figure 7). By using the lower F max , a higher cyclic hardening potential at the crack edges in relation to the plastically deformed volume at the crack tip can be observed. This correlates to sub-grain formation at the crack edges caused by plastic deformation during crack propagation, which is less pronounced in the plastic zone around the crack tip (see Figure 8). However, this cannot explain the decreased hardness in the vicinity of crack edges compared to the crack tip. Moreover, it has to be considered, that the areas at the crack edges were plastically deformed in an earlier state of the TMF-test. This increases the possibility of the Laves phase particle formation at dislocations, which probably also influences cyclic hardening potential and hardness. This will be the focus of the ongoing work. To analyze the differences between the crack tip and crack edge, CITs with Fmax = 100 mN were conducted, leading to a higher spatial resolution. Due to the indentation size effect (see References [16,22]), the CITs with Fmax = 100 mN yield a higher hardness compared to Fmax = 1000 mN (see Figure  7). By using the lower Fmax, a higher cyclic hardening potential at the crack edges in relation to the plastically deformed volume at the crack tip can be observed. This correlates to sub-grain formation at the crack edges caused by plastic deformation during crack propagation, which is less pronounced in the plastic zone around the crack tip (see Figure 8). However, this cannot explain the decreased hardness in the vicinity of crack edges compared to the crack tip. Moreover, it has to be considered, that the areas at the crack edges were plastically deformed in an earlier state of the TMF-test. This increases the possibility of the Laves phase particle formation at dislocations, which probably also influences cyclic hardening potential and hardness. This will be the focus of the ongoing work.   To analyze the differences between the crack tip and crack edge, CITs with Fmax = 100 mN were conducted, leading to a higher spatial resolution. Due to the indentation size effect (see References [16,22]), the CITs with Fmax = 100 mN yield a higher hardness compared to Fmax = 1000 mN (see Figure  7). By using the lower Fmax, a higher cyclic hardening potential at the crack edges in relation to the plastically deformed volume at the crack tip can be observed. This correlates to sub-grain formation at the crack edges caused by plastic deformation during crack propagation, which is less pronounced in the plastic zone around the crack tip (see Figure 8). However, this cannot explain the decreased hardness in the vicinity of crack edges compared to the crack tip. Moreover, it has to be considered, that the areas at the crack edges were plastically deformed in an earlier state of the TMF-test. This increases the possibility of the Laves phase particle formation at dislocations, which probably also influences cyclic hardening potential and hardness. This will be the focus of the ongoing work.  Conclusively, the presented investigations at the fatigue crack demonstrate, that CITs can be used to determine local differences in the material's cyclic properties, which correlates to microstructural changes observed. The evolution of the complex microstructure of Crofer ® 22 H in the fatigue crack vicinity, and its influence on the results obtained in CITs will be investigated intensively in future work.

Conclusions
In the presented work, the damage tolerance of 22 wt.% Cr, fully ferritic stainless steel Crofer ® 22 H, as well as advanced ferritic-martensitic (AFM) steel P91, was analyzed using the short-time procedure PhyBaL CHT . This was based on instrumented cyclic indentation tests (CITs), which were performed at specimens and loaded for the different numbers of oop TMF-cycles, as well as in the vicinity of a fatigue crack in a Crofer ® 22 H specimen. Consequently, the presented work focuses on the evolution of damage tolerance during TMF-loading, as well as the local changes in defect tolerance in the vicinity of a fatigue crack. The following conclusions can be drawn from the results obtained:

•
In the initial state, Crofer ® 22 H has a higher damage tolerance compared to P91, which correlates to higher TMF-lifetime;

•
After approximately half of TMF-lifetime, an increase of damage tolerance can be observed for Crofer ® 22 H, which is not seen for P91 and correlates with changes of the Laves phase particle size and distribution; • Polygonization, resulting from degradation of martensite lath structure in P91, does not influence the microhardness, as well as cyclic hardening potential obtained in CITs; • For Crofer ® 22 H, the changes in damage tolerance correlate well with the cyclic deformation behavior observed in TMF-experiment and the accompanied microstructural changes; • The cyclic hardening potential, and hence, damage tolerance, as well as hardness, is different between the crack tip and crack edges in Crofer ® 22 H; • CITs are a powerful means to analyze the TMF-induced changes in mechanical properties, i.e., cyclic hardening potential, as well as for the investigation of locally restricted changes in the material's cyclic deformation behavior.
However, the presented work also shows that a thorough understanding between the microstructural changes (Laves phase particle evolution, dislocation density, and sub-grain boundary formation) and the resulting deformation behavior observed in CITs is a prerequisite. This will be considered in further investigations.

Funding:
The financial support of the priority research activity of Rhineland Palatinate "Advanced Materials Engineering (AME)" is gratefully acknowledged.