Creep strength boosted by a high-density of stable nanoprecipitates in high-chromium steels

Abstract There is a need worldwide to develop materials for advanced power plants with steam temperatures of 700°C and above that will achieve long-term creep-rupture strength and low CO2 emissions. The creep resistance of actual 9-12Cr steels is not enough to fulfil the engineering requirements above 600°C. In this paper, the authors report their advances in the improvement of creep properties of this type of steels by the microstructural optimization through nano-precipitation using two methodologies. 1) Applying a high temperature austenitization cycle followed by an ausforming step (thermomechanical treatment, TMT) to G91 steel, to increase the martensite dislocation density and, thus, the number density of MX precipitates (M = V,Nb; X = C,N) but at the expense of deteriorating the ductility. 2) Compositional adjustments, guided by computational thermodynamics, combined with a conventional heat treatment (no TMT), to design novel steels with a good ductility while still possessing a high number density of MX precipitates, similar to the one obtained after the TMT in G91. The microstructures have been characterized by optical, scanning and transmission electron microscopy, EBSD and atom probe tomography. The creep behaviour at 700°C has been evaluated under a load of 200 N using small punch creep tests.


Introduction
The need to look for ways to ensure the sustainability of current energy sources to guarantee the viability of future generations, as long as they are environmentally friendly, is beyond dispute.The future of power plants is to reduce fuel costs and CO 2 emissions thought improvements in their efficiency by elevating the steam conditions to even higher ranges of pressure and temperature than those applied nowadays.
The essential function of a power station is to produce electrical energy using different sources (fossil fuel, nuclear fuel, wind, solar…etc).Regarding nuclear/fossil fuel power plants, investigations focus on developing structural materials that withstand higher steam temperatures.
As it has been reviewed by Klueh in his seminal work on high-Cr ferritic-martensitic (FM) steels (Klueh & Harries, 2001), the design and production of 9-12Cr FM steels began in 1912 when Krupp and Mannesmann produced a 12 wt.%Cr steel containing 2-5 wt-% Mo.The zeroth generation that contained mainly 9-12 wt.% Cr evolved to the 12CrMoV steels introduced in the power plants in the mid 1960s for thin and thick-walled power station components.Their creep strength was based on the solid solution hardening of substitutional elements and on the precipitation of M 23 C 6 carbides.These steels have been applied successfully in power stations over several decades (Masuyama, 2001).These steels had 105 h rupture strengths at 600°C under a stress of up to 60 MPa.
The second generation, developed in the late 1970s, is based in the modified 9Cr-1Mo, designated as G91 and HCM12, which were developed for the manufacturing of pipes and vessels for fast breeder reactors (Masuyama, 2001).In these steel class, the balance between C, Nb and V was optimized, N (0.03-0.05 wt.%) was added, and the maximum operating temperature increased to 593°C.The new steels have a duplex structure (tempered martensite and δ-ferrite).These steels have 105 h rupture strengths at 600°C under a stress of about 100 MPa.Of these latter steels, G91 has been used most extensively in the power-generation industry in all new power plants with steam temperatures up to 600°C (Klueh & Harries, 2001).
In the early 1990s, a Japanese steel development program of Nippon Steel led to the development of the P92 steel (NF616).With the P92, designated Grade 92, a further increase in creep strength was obtained by the addition of 0.003 wt.% B, 1.8 wt.% W and a reduction of the Mo content from 1 to 0.5 wt.% (Ennis & Czyrska-Filemonowicz, 2003;Wasilkowska, Bartsch, Messerschmidt, Herzog, & Czyrska-Filemonowicz, 2003).The addition of B ensures thermally stable M 23 (C,B) 6 precipitates, whereas the higher W content leads to higher amounts of precipitated Laves phase (Abe, 2008;Eggeler & Dlouhy, 2005).The Grade 92 achieves an operation temperature of 620°C.
Finally, the next generation of steels is being developed at present, where the intention is to push the operating temperatures to 650°C or above.These fourth-generation steels differ from the previous generation primarily by the use of up to 3.0 wt.% Co as an austenite stabilizer because of the adverse effect of Ni on creep (Klueh & Harries, 2001).In these steels with about 0.1 wt.% C, Mo has been further reduced or eliminated, and W (2.6-3.0 wt.%) has been increased compared to the third-generation compositions.
The common denominator of all steel developments for high temperature components in power plants is to combine properties such as fabricability, corrosion resistant and creep strength.As it was indicated above, this last property is the most important feature and it has led innumerable research activities aiming at improving the creep strength in 9Cr FM steel developments (Pasternak & Dobrzanski, 2011;Hald, 2008;Skelton & Gandy, 2008;Skelton, 1993;Cane, 1996).The main shortcoming for high-temperature application of these steels is their loss of strength beyond 600°C.This review compiles the research carried out in our research group in the last years to achieve a microstructural stability at elevated temperatures in 9Cr FM steels by applying a combination of compositional design and thermomechanical treatments (TMT), instead of a conventional treatment, triggering the precipitation of MX (M = Nb,V; X = C,N) nanosized precipitates during the tempering stage.Therefore, the originality of this research line consists on demonstrating that by combining compositional adjustments and thermomechanical control process, a thermal stable microstructure consisting on nanosized particles distributed in a number density equivalent to those more costly produced oxide dispersion alloys is feasible.

Experimental techniques
The thermomechanical simulations were carried out in a 805 DIL Bahr plastodilatometer using cylindrical samples of different dimensions; some for the microstructural examination (10 mm in length and 5 mm in diameter) and larger ones for Small Punch Creep (SPC) tests (10 mm in length, 8 mm in diameter).The samples were heated at 5°C/s and cooled at 50°C/s.The deformation applied was 20% at 0.1 s −1 .Figure 1 shows two types of processing routes: the first one emulates the industrial manufacturing conditions for the G91 steel and has been named as AR (as-received) process.A similar processing route has been applied also to the HDSN steels that will be introduced in section 4 (so-called HDSN process in Figure 1).The second route is a thermomechanical treatment (TMT process) where the combined effect of increasing the austenitization temperature and applying deformation will be studied.It is worth mentioning that the austenitization temperature must be low enough to avoid the presence of delta ferrite because of its detrimental effect on long-term creep properties.The optimum austenitization temperature of 1225°C was calculated by means of Thermocalc® (Vivas, Capdevila, Jimenez, Benito-Alfonso, & San-Martin, 2017).
Samples were prepared by conventional metallographic techniques for microstructural examination by light optical microscopy (LOM) and scanning electron microscopy (SEM).Transmission electron microscopy (TEM: JEOL JEM 2100 and JEOL JEM 3000 F) was used to observe the precipitates present in the microstructure and determine the size and number density.TEM samples were prepared by electropolishing (Tenupol 5) with a 95/5 acetic/perchloric acid solution at 25°C and 40 V.
Quantitative X-Ray diffraction analysis was performed on polished samples with 50 nm colloidal silica suspension.X-Ray diffraction measurements were performed by means of a Bruker AXS D8 diffractometer equipped with a Co-Kα X-ray tube with a Goebel mirror optics and a LynxEye Linear Position Sensitive Detector for ultra-fast XRD measurements.For the Rietveld refinement of the diffractograms, the version 4.2 program TOPAS has been used and the crystallographic information of the phase was obtained from Pearson´s Crystal Structure Database for Inorganic Compounds.The contribution of the microstrain is evaluated from the width of the diffraction peaks by means of the Warren-Averbach method (Balzar & Ledbetter, 1993).Once the microstrain is determined, the dislocation density is evaluated taking the most accepted formula in the literature that relates both (Miyamoto, Iwata, Takayama, & Furuhara, 2012;Tan, Snead, & Katoh, 2016).
For the atom probe tomography (APT) measurements, several needles were prepared from each steel composition.The APT experiments were performed in a local electrode atom probe (CAMEACA LEAP 4000X HR).The APT needles were prepared by means of the well-established FIB-based site-selective lift-out method to pinpoint locations inside martensite laths away from boundaries.Sections of the lift-out were mounted onto a Si microtip array, sharpened using annular milling, and cleaned with a 2 kV Ga + ion beam (Thompson et al., 2007).The APT needles were run in laser mode with a 50 pJ laser pulse energy, 200 kHz pulse repetition rate, a -243°C base temperature, and a 0.5% detection rate.CAMECA's Integrated Visualization and Analysis Software (IVAS) version 3.8.2 was used for the 3 D reconstructions and data analysis.
Due to the limited amount of material available after the thermal and thermomechanical treatments carried out in the plastodilatometer, the creep properties were investigated by means of SPC tests performed at 700°C.The SPC samples were cut transversally, from cylindrical specimens, with a thickness of 600 µm and a diameter of 8 mm.Then, the discs were ground on both sides down to a final thickness of 500 µm.Figure 2 shows the set-up of the SPC tests and it is described elsewhere (Vivas et al., 2018).The minimum disk deflection rate (δd) depends on the applied force (F) by the Norton Law (Norton, 1929) rewritten in the following way:

Effect of thermomechanical control processing on the microstructure
The G91 steel is used as a benchmark for the development of steels with upper-use temperatures of 600-620°C.However, it is difficult to keep pushing to higher operating temperatures too much.It is required to develop a method that increases the number density of thermally stable nanosized particles and promotes them homogeneously distributed in the matrix, i.e. the oxide dispersion strengthened (ODS) steels (Benjamin, 1970;Klueh, Shingledecker, Swindeman, & Hoelzer, 2005) were born.
The ODS steels are strong and stable at high temperatures, but the complicated and expensive powder metallurgy based manufacturing route avoided their full implantation as structural material for large components in the current power plants.Therefore, an alternative strategy to achieve a high number density of nanoprecipitates is needed.In this section, we present preliminary results that allow us to conclude that conventional thermomechanical control processing strategies are adequate to achieve nanosized particle dispersion-strengthened steels.The material considered in this study is the commercial G91 bearing in mind that a high number density of fine MX precipitates (Nb-MX and V-MX) triggered by a controlled TMT should display superior high temperature performance.

Microstructure after the conventional heat treatment
The manufacturing process of the commercial G91 steel (AR process in Figure 1) consists of an austenitizing step at 1040°C for 30 min, followed by air quenching, and finally tempering at 730°C for 1 hour as indicated in Figure 1.The microstructure consists of tempered martensite, which presents elongated subgrains with an average size of 0.25-0.50µm (Figure 3).Two types of precipitates, M 23 C 6 carbides rich in Cr and MX carbonitrides rich in V or Nb are present in the microstructure.The size of M 23 C 6 carbides is around 100-200 nm and they are precipitated on subgrain boundaries and prior austenitic grain boundaries.The size of MX carbonitrides is much smaller than the M 23 C 6 carbides, 20-50 nm, and they nucleate intragranularly in the matrix (Klueh, 2005).
Lath martensite is a particular microstructure that ensures microstructural stability.Furuhara and Miyamoto (Miyamoto, Iwata, Takayama, & Furuhara, 2013;Miyamoto et al., 2012) described the variety of crystalline sizes in lath martensite structures.A hierarchy of lath martensite structure is clearly identified particularly in low carbon steels.A prior austenite (γ) grain is divided into "packets" each of which consists of a group of martensite laths with the same parallel close-packed plane relationship in the Kurdjumov-Sachs (K-S) orientation relationship denoted as 'CP group' recently.In general, a packet is partitioned into several blocks, each of which contains laths of a single variant of the K-S relationship.Blocks and packets are mostly surrounded by high-angle boundaries whereas lath boundaries inside a block are of low-angle type.

Effect of the ausforming
Several works report the role of austenite deformation on refining the martensitic microstructure (Chiba, Miyamoto, & Furuhara, 2012;Miyamoto, Iwata, Takayama, & Furuhara, 2010).Deformation temperatures above the austenite recrystallization temperature can produce an austenite grain refinement that can induce the concomitant martensitic microstructural refinement.Similarly, plastic deformation below the recrystallization temperature, the so-called ausforming (Tamura, Tsuzaki, & Maki, 1982), promotes the formation of deformation bands in the austenite which directly induce that some specific martensite crystallographic variants are preferentially formed upon quenching, leading to the development of a strong transformation texture (martensite variant selection) and anisotropic properties.
The effect of thermomechanical treatments on tempered martensite is studied on TMT cycles presented in Figure 1.The next stage after the austenitization is the ausforming.In this stage, austenite is deformed to introduce dislocations which will act as nucleation sites for precipitates in the subsequent tempering.Miyamoto et al (Miyamoto et al., 2012) reported a detailed description of the effect of ausforming on low-carbon lath martensite.The authors described that martensite variants with habit planes, i.e. (575) γ , that are nearly parallel to the close-packed primary and secondary slip planes in austenite, i.e. (111) γ and (-111) γ , nucleate and grow preferentially.Therefore, the strain accumulated during the ausforming results in an increasing number of dislocations in the (111) γ and (-111) γ slip planes that might be transferred to the martensite (011) M planes.In this sense, ausforming might increase the dislocation density in the martensite.
Two different ausforming temperatures were studied, 600 and 900°C, for a same level of plastic deformation of 20%.The dislocation densities after each ausforming process are 2.8 ± 0.1 × 10 15 m −2 and 1.9 ± 0.1 × 10 15 m −2 , respectively, determined form XRD measurement of the microstrain as described in previous section.These results are consistent with the broad idea that the lower ausforming temperature, the higher dislocation density introduced in austenite is.This increase in the dislocation density in fresh martensite with a decrease in the ausforming temperature is explained, on the one hand, because some of the dislocations are inherited from the deformed austenite as it was explained in this work previously.On the other hand, the deformation of austenite refines lath thickness and reduces martensite start temperature as Zhang et al. reported in a previous work (Zhang, Wang, Zheng, Zhang, & Wang, 2014).
During the final stage (tempering), MX precipitates and M 23 C 6 carbides precipitate and the recovery of dislocations takes place.Figure 4 illustrates the microstructures after the TMT process compared to the AR one.This figure presents IPF-EBSD maps, SEM and TEM micrographs of the martensite matrix, M 23 C 6 precipitates and MX nanoprecipitates distributions after the TMT and AR processes in steel G91.The APT analysis in this figure show 2.5 at.%V isoconcentration surfaces and the corresponding 1 D concentration profile along the nanoparticle denoted by a black arrow in the isoconcentration surfaces.However, the APT is not directly correlated to the TEM image though they both refer to the same heat treatment condition (ausforming at 900°C).The Fe concentration in the precipitated, depicted in the 1 D profile, is due to the trajectory aberrations and ion crossing in the APT dataset, which is caused by the irregular shape of the APT needle from different evaporation fields between the matrix and the precipitates.The composition overlapping between the matrix and the precipitate is enhanced as the precipitate size decreases.The microstructure of steel G91 after the TMT and AR processing treatments consists of tempered martensitic laths with M 23 C 6 carbides on lath, block and prior austenite grain boundaries.Finer MX precipitates are found homogeneously distributed within laths.The APT analysis reported in (Vivas, Capdevila, et al., 2019) allows us to conclude that some of MX precipitates are located on dislocations, which could demonstrate the potential of dislocations to act as nucleation sites for these precipitates.Besides, the nature of those MX nanosized precipitates is (V,Nb)(C,N) as reported by Vivas et al (Vivas, Capdevila, et al., 2019).
The role of dislocations on the precipitation process during tempering was recently illustrated by Takahashi et al.These authors recently reported the formation of Nb-Cottrell atmospheres in low-carbon Nb-microalloyed steels since the segregation energy of Nb to the edge dislocation core was almost the same as the grain boundary segregation energy, and the large attractive interaction between the Nb and the dislocation was due to its large atomic size.Such interaction Nb-dislocation retards the recovery of the dislocations at high temperatures (Takahashi, Kawakami, Hamada, & Kimura, 2016).In the steel investigated it could be expected that Nb behaves in a similar way, retarding the recovery after ausforming and promoting the precipitation of fine and homogeneous MX carbonitrides during tempering.
Vivas et al. identified different stages concerning the formation process of MX nanoparticles (Vivas, Capdevila, et al., 2019).After quenching, and before tempering, only C atoms appear segregated to martensitic defects.In the early stage of tempering V, Nb, Cr and N segregate to dislocations, which proves that dislocations act as nucleation sites for these nanoprecipitates.As the tempering time increases, nanoparticles (5-10 nm in size) enriched in Nb, V, N, Cr and Fe are observed.Finally, at the end of tempering, nanoparticles enrich on Nb, V, Cr, N and lose Fe.
So far, it was described the effect of ausforming on promoting dislocations that act as potential nucleation sites for MX-precipitation during tempering stage.However, it is crucial to determine the effect precipitates and MX nanoprecipitates distributions after the tMt and ar processes in steel G91. the apt analysis in this figure show 2.5 at.%V isoconcentration surfaces and the corresponding 1 D concentration profile along the MX nanoparticle denoted by a black arrow in the isoconcentration surfaces.the apt is not directly correlated to the tEM image though they both refer to the same heat treatment condition (ausforming at 900°C).after (Vivas, Capdevila, et al., 2019) of ausforming on the number density of nano sized MX precipitates.In this sense, the number density of MX precipitates (N) was determined through direct measurements of the spacing (λ), between MX carbonitrides, from several TEM micrographs, and following Eq.( 2) (Futamura, Tsuchiyama, & Takaki, 2001) and stereology principles (Russ, 1986): The size of MX carbonitrides was 5.6 nm and 7.4 nm for the materials ausformed at 600°C and 900°C, respectively.Therefore, the number density of MX carbonitrides was 9.39 × 10 22 m −3 and 6.4 × 10 22 m −3 for the material ausformed at 600 and 900°C, respectively.The reported values of size and number density of MX carbonitrides after conventional processing were 30 nm and 10 20 m −3 , respectively (Tan et al., 2016).Therefore, after the ausforming, precipitates size is reduced up to five times and the number density is increased up to two orders of magnitude.In fact, these number densities and precipitates sizes are very similar to those corresponding to oxides in oxide dispersion strengthened (ODS) steels (Toualbi et al., 2012;Heintze et al., 2011).ing Eq.( 1).The value of R 2 =0.98 in the latter suggests that all the samples tested obey the same creep mechanism.Thus, from this result it can be deduced that TMT processing does not affect the creep mechanism.The characteristic value of the exponent n for the materials is found to be 5.5.A value of 6 for the stress exponent was referenced by Abe et al. (Abe, 2015) in G91 steels at low stresses.The value of 5.5 implies that creep deformation mechanism is controlled by the dislocation motion.It should, however, be noted that the exponent n is not exactly the same as the one obtained from uniaxial creep tests.In the SPC tests, the equivalent stress is not homogeneous within the sample.Moreover, the stress level is increasing with the increasing deflection (time).Nevertheless, it is meaningful to compare the load exponent of various SPC tests.

Creep behaviour
It is worth noting that the samples after TMT processing exhibit higher creep rupture strength than after the AR condition, which allow us to conclude that the significant increase in the number density of nano sized MX precipitates plays the role.It is also interesting noting the significant difference between the TMT processed steels.A higher creep rupture strength is obtained for the sample processed at the lowest ausforming temperature.The same trend was observed with the minimum disk deflection rate, which indicates that the pinning effect of MX precipitates on dislocations during creep allows retarding the onset of acceleration creep, increasing considerably the time to rupture.

Post-creep analysis
Figure 6 provides scanning electron micrographs of the fractured specimens after the SPC tests for the AR and TMT samples tested.The AR processed sample exhibits a significant reduction in thickness and no radial cracks have been observed (Figure 6c) suggesting a ductile fracture behaviour.By contrast, TMT processed samples (Figure 6a and b) shows the presence of radial cracks indicating a brittle fracture.It is important to mention that no significant differences in ductility have been observed among the TMT samples.
Forensic analysis of the SPC tested samples allows us to clarify the failure mechanisms.The light optical images provided in Figure 7 indicate that the AR sample exhibits the presence of transgranular cavities (Figure 7a).By contrast, the TMT samples show the cavities growing mainly along the prior austenite grain boundaries.This fact suggests that the degradation of the microstructure occurs heterogeneously and starts at the vicinity of the prior austenite grain boundaries on cavities attached to coarse particles (Figure 7c).By contrast, the presence of transgranular cavities in the AR samples suggests that during the SPC tests the for the ar and tMt processed samples.after (Vivas et al., 2018).
degradation of the microstructure happens homogeneously at random locations in the matrix.
The coarse M 23 C 6 carbides located at the prior austenite grain boundaries contribute to the inhomogeneous and localized deformation experienced by the TMT samples during creep.The local creep concentration close to the prior austenite grain boundaries would be promoting the nucleation of cavities.Besides, in the TMT samples, the high austenitization temperature produces an enormous prior austenite grain size with the concomitant large grain boundary surfaces, facilitating an earlier formation of the critical crack length that lead to the intergranular fracture, with the brittle behaviour shown in Figure 7 (Plesiutschnig, Beal, Paul, Zeiler, & Sommitsch, 2015;Anderson, 2017).

Novel highly reinforced heat resistant steels
So far, the feasibility of processing a commercial grade such as G91 to boost the number density of nanoprecipitates has been shown.The strategy consisting of combining an increase of the austenitization temperature and the ausforming processing, rose the number density of nanosized MX precipitates up to 3 order of magnitude as compare with the conventional processing route.But at a cost.The inherent coarse prior austenite grain promotes a brittle failure that must be avoided.In this sense, this section explores an alternative strategy based on the optimization of the MX nanoprecipitates distribution but bearing in mind that the prior austenite grain size must be controlled during the processing route to avoid the loss in creep ductility.

Steel design
Three new steels were designed to be an evolution of the commercial G92 and G91 steels but presenting a higher amount of nanoprecipitates  (Vivas, Capdevila, et al., 2019).
that implement the strengthening of the steel at high temperatures.The three of them have been designed to develop a Fe-BCC matrix but reinforced by nanosized precipitates.Different strategies were implemented to achieve this goal of controlling the chemistry of the ferritic matrix and the chemistry and location of the precipitates.The resulting High-Density of Stable Nanoprecipitates (HDSN) steels are listed in Table 1.
In this sense, the improvement in the oxidation behaviour at high temperature (650-700°C) is expected to be achieved by the addition of 9 wt.%Cr.The required solid solution strengthening is expected to be accomplished by the addition of W and Mo, which will lead to higher creep strength level at the expected operating temperatures.Besides, these two elements retard the recovery of martensitic laths by the sluggish self-diffusion rate (Abe, Horiuchi, Taneike, & Sawada, 2004).These elements also lower the Ms Temperature, which would increase the dislocation density of martensite (Nedjad, Moghaddam, Vazirabadi, Shirazi, & Ahmadabadi, 2011).By contrast, W and Mo promote the Laves phase formation, which presents a poor thermal stability (Wang et al., 2013;Hu et al., 2010).Therefore, HDSN1 steel was designed with addition of Mo and W and only Mo for HDSN2 steel.
Moreover, a proper balance between interstitial solutes (C and N) and carbonitride formers (V and Nb) is required.The V and Nb are strong carbide formers that promote the MX precipitation that will pin the dislocations during creep and extend the time to rupture (Hasegawa, Tomita, & Kohyama, 1998;Liu, Zhang, Xia, & Yang, 2014).Similarly, a proper content of C and N was added to achieve precipitation strengthening by the formation of M 23 C 6 carbides and MX precipitates.High contents of C and N must be avoided to facilitate the weldability (Altstadt, Serrano, Houska, & García-Junceda, 2016).Finally, the addition of Co and Ni expands the austenite field and avoid the formation of delta ferrite during the austenitization heat treatment (Helis, Toda, Hara, Miyazaki, & Abe, 2009).after (Vivas, Capdevila, et al., 2019).
In contrast to HDSN1 and HDSN2 steels, the HDSN3 steel was designed to present a higher population of MX nanosized nitrides.To achieve this, the C content was reduced to 0.015 wt.% and the N content increased up to 0.098 wt.%.In addition, this reduction in C leads to a considerable reduction of the amount of M 23 C 6 carbides.It is worth noting that Nb was not added to this steel and the MX precipitates expected to be formed are V-type precipitates in contrast to the previous designs, i.e.HDSN1 and 2 steels, where MX precipitates were designed to be (V,Nb)(C,N).Furthermore, as in the HDSN1, W for solid solution strengthening and Co for avoiding delta ferrite formation were added.The higher W content for this steel, compared to the HDSN1, would promote the formation of the Laves Phase at the tempering temperature instead of M 23 C 6 carbides.

Microstructural determination of key-parameters for enhanced creep performance
The steels listed in Table 2 were produced by vacuum induction melting.The ingots were subjected to a hot-forging of 50% at 1200°C and air cooled to room temperature.Subsequently, they were subjected to the thermal cycle described in Figure 1, i.e. they were austenitized for 10 minutes at 1050°C followed by air cooling down to room temperature, then tempered for 1 hour at 730°C and air cooled to room temperature.It is worth noting that this thermal processing involves heat treatments without the ausforming stage, in contrast to the TMT processed G91 samples studied in sections above.However, the dislocation densities obtained in these steels are similar or greater than for the ausformed conditions in G91: HDSN1 (2.3 × 10 15 m −2 ); HDSN2 (2.1 × 10 15 m −2 ) and HDSN3 (3.3 × 10 15 m −2 ) (Liu et al., 2014).This high density along with the high amounts of MX forming elements contribute to having a high number density of MX nanoprecipitates similar to the ones observed in the ausformed samples of G91.
Figure 8 shows TEM micrographs for HDSN1 and HDSN2 steels to disclose the distribution of the MX nanoprecipitates (Vivas et al., 2020).Similar to the situation found in the commercial G91 grades, the M 23 C 6 carbides are located on block and lath boundaries, and nanosized MX precipitates appear located within the laths.Regarding HDSN3 steel, fine  (c)n HDsn3 after the tempering treatment.the tEM images and the apt results are not correlated but do refer to the same heat treatment condition.after (Vivas et al., 2020;Vivas et al., 2021).
MX nanoprecipitates are also observed within the laths (Figure 8c) in a high number density.It is also worth mentioning that while the size of MX precipitates within the laths are ranging from 10-30 nm, those located on prior austenite, block and lath boundaries are larger than 100 nm.Vivas et al. identified the nature of the MX-nanoprecipitates in HDSNs samples (Vivas, Poplawsky, De-Castro, San-Martín, & Capdevila, 2021).Figure 8 shows the 1 at.%V isoconcentration surface (right side reconstruction), and the 1 D concentration profile (middle plots) of elements Cr, V, W, Co, C and N, performed along the nanoprecipitate denoted by the black arrow in the isoconcentration surface for the three HDNS steels investigated.The TEM was employed to estimate the average size and number density of the MX precipitates while the APT was used to investigate their composition.The TEM and APT results in this figure are not correlated but refer to the same heat treatment condition.The conducted APT measurements (Figure 8) showed that the cores of the detected nanoprecipitates are enriched in V, Nb, Cr and N in the HDSN1 and HDSN2 steels.For the case of the HDSN3 steel, the cores of the nanoprecipitates contained high amounts of Cr, V and N.These results indicate that all these precipitates are nitrides.Other elements like Co or W appear homogeneously distributed within the martensitic microstructure; they did not partition to precipitates nor segregated to dislocations.Regarding the size of the nanoprecipitates, no important differences were observed regardless of the HDSN steel considered.These Figure 9. small punch creep test (spCt) curves obtained for the HDsns, G91-ar and G91-tMt 600 steel samples at 700°C (load = 275 n). after (Vivas et al., 2020).
results highlight the importance of V, Nb, Cr and N as alloying elements to form nanoprecipitates that boost the creep strength of these heat resistant steels.
On the other hand, Table 2 lists the number density and average particle size of M 23 C 6 carbides and MX nanoprecipitates in HDSN1-3.HDSN1 and HDSN2 steels display very similar number density and average particle size for M 23 C 6 carbides and MX nanoprecipitates.However, the number density of MX nanoprecipitates for HDSN3 steel is smaller than that measured for HDSN1 and HDSN2 steels.This might be explained by the differences in size of the MX precipitates since different nucleation sites in the HDSN1 and HDSN2 steels as compared with HDSN3 steel.Meanwhile, MX precipitates appear within the laths in the HDSN1 and HDSN2 steels; additionally, they also do on block and prior austenite grain boundaries in the HDSN3 steel, due to the absence of M 23 C 6 carbides at these latter locations.These locations for nucleation lead to coarser precipitates.

Creep behaviour: Small punch creep tests
Figure 9 depicts the punch displacement vs. time curves for the HDSN steels at 700°C and 275 N.For the sake of comparison, this figure also includes the SPC curves for G91 samples processed under AR (G91-AR) and TMT at 600°C (G91-TMT 600) conditions.From these results one might conclude that there is no significant difference among the HDSNs and G91-TMT 600 samples.However, an important decrease in the minimum displacement rate and a great increase in the time to rupture are observed among the HDSNs steels compared to those observed for the G91-AR steel sample, which reveal the creep strengthening ability of the new designed steel as compared with the commercial existing ones.In this sense, values of minimum displacement rate of 5.81, 7.72 and 6.02 µm h −1 for HDSN1, HDSN2 and HDSN3 samples were recorded, meanwhile values of 19.56 µm h −1 for G91-AR and 2.7 µm h −1 for G91-TMT 600 samples were recorded.Regarding the time to rupture values of 50.02, 44.65 and 39.37 h for HDSN1, HDSN2 and HDSN3 samples were recorded, meanwhile values of 19.28 for G91-AR and 49.11 h for G91-TMT 600 samples were recorded.
The HDSNs samples exhibited similar minimum displacement rate and almost the same time-to-rupture values than the TMT processed samples of the G91 steel.The creep behaviour of HDSN1 and HDSN2 steels might be explained on the same basis sustained for the TMT samples of G91 grade in sections above, i.e. by the high number density of MX nanoprecipitates.The HDSN1 and HDSN2 samples present the same values in terms of number density and size of precipitates than those measured for the TMT samples at 600°C for the G91 steel.These MX nanoprecipitates are the most numerous obstacles to dislocation motion and, in consequence, control the recovery of the dislocation substructure.Thus, similar minimum disk deflection rate might be expected.Particular attention should be paid to the case of the HDSN3 samples.As mentioned above, the number density of the MX precipitates is significantly lower than for the HDSN 1 & 2 samples.Thus, the improvement in creep strength cannot be explained just considering the MX distribution within the laths.The improvement in creep strength might be justified by the increase in solid solution strengthening induced by the higher concentration of W in the chemistry, which is consistent with the work of Taneike et al (Taneike, Abe, & Sawada, 2003).
It is important to consider that the creep tests duration is very short and only reflects the influence of the initial microstructures produced by the processing route.Iseda et al [56] demonstrated that dislocation strengthening contributes greatly to extend the time to rupture in short time creep tests.However, this strengthening becomes useless during long-term creep because of the annihilation of dislocations during the recovery and the recrystallization that takes place during the creep test.Nevertheless, Maruyama et al. (Maruyama, Sawada, & Koike, 2001) suggested that a high dislocation density can be useful if the premature recovery of dislocation substructure is prevented by some means.In the case of HDSN steels, the approach used to retard the recovery and recrystallization is by promoting a homogeneous and fine distribution of MX precipitates in the microstructure, which are able to pin the dislocations during creep.Hence, these precipitates would extend the strengthening contribution of the dislocations to longer times and, consequently, promote a high creep strength even in long-term creep tests.
The fracture surface of the HDSN1 (Figure 10a) and HDSN2 (Figure 10b) shows a dimple pattern for both steels, which is characteristic of ductile transgranular fractures.By contrast, a clear cleavage fracture characteristic of brittle fractures was observed for the TMT processed samples at 600°C of the G91 steel (Figure 10c).The ductile fracture mechanism for the HDSN steels is caused by the fine prior austenite grain size resulting from the low austenitization temperature.The relatively low austenitization temperature of 1050°C allows limiting an excessive grain growth, which was the main factor that triggered the brittle fracture in the TMT processed samples of the G91 steel.Besides, the grain growth was inhibited by the grain boundary pinning effect provided by the MX precipitates present during the austenitization (primary carbides).

Effect of ausforming on creep behaviour
The combination of increasing austenitization temperature from 1040 to 1225°C and ausforming processing at 900°C allow increasing number density of MX up to 3 orders of magnitude which raises the strengthening capability of MX at 700°C up to 6.5 times.In fact, the number density of nano-sized precipitates is in the same order of magnitude than those measured in ODS steels.Ausforming increased the dislocation density in martensite, that might boost the potential nucleation sites for MX nanosized precipitation of further tempering.This type of microstructures reduced considerably the minimum disk deflection rate and showed greater time to rupture during the SPC tests carried out at 700°C.By contrast, the elevated austenitization temperature required to drive into solid solution most of potential carbide formers induces an important drop in ductility.This drop in creep ductility of the G91 steel after the TMT processing is a consequence of the preferential formation of cavities close to coarse M 23 C 6 carbides located at the vicinity of prior austenite grain boundaries.

Novel high-density of stable nanoprecipitates steels
Computational metallurgy techniques have been used to design a high density of stable MX nanoprecipitates to improve the creep strength of the commercial G91 steel grade.The amount of MX precipitates was tailored using thermodynamic calculations supported by Thermo-Calc® and designed to be in the range around 0.6-0.85mol %, which is considerably higher than the amount (0.4 mol %) predicted for conventional 9Cr FM steels.The number density of MX nanoprecipitates measured in the microstructure obtained was ranging between ∼10 20 and 10 22 m −3 for the HDSN steels, which is comparable to the TMT processed samples of the G91 grade, and substantially higher than the commercially produced G91 steel (∼10 19 m −3 ).(Vivas et al., 2020).
The SPC results at 700°C indicate the same level than the TMT processed samples of the G91 steel, and a remarkable improvement compared with the commercial G91 steel grade was found.Besides, the relatively fine austenite grain size achieved in the HDSN steels promotes a significant increase in creep ductility as compared with the TMT processed samples.

Figure 1 .
Figure 1.scheme of the thermal and thermo-mechanical treatments applied in this work to the G91 steel samples (ar and tMt processes) and HDsn steels (HDsn process).
where δd min is the minimum disk deflection rate obtained at the load tested for each processing condition.the representation of Ln mechanism through the value achieved for the n-exponent (slope).

Figure 3 .
Figure 3. (a) representation of the resulting hierarchy microstructure in the G91 steel achieved by the ar process shown in figure 1; (b) and (c) sEM micrographs showing coarse, white M 23 C 6 precipitates are located at lath, block and prior austenite grain boundaries; (d) and (e) tEM micrographs showing the presence of MX nanoprecipitates within the laths and coarse M 23 C 6 at a lath grain boundary.

Figure 4 .
Figure 4. ipf-EBsD maps, sEM and tEM micrographs of the martensite matrix, M 23 C 6precipitates and MX nanoprecipitates distributions after the tMt and ar processes in steel G91. the apt analysis in this figure show 2.5 at.%V isoconcentration surfaces and the corresponding 1 D concentration profile along the MX nanoparticle denoted by a black arrow in the isoconcentration surfaces.the apt is not directly correlated to the tEM image though they both refer to the same heat treatment condition (ausforming at 900°C).after(Vivas, Capdevila, et al., 2019)

Figure 5
Figure 5 shows the SPC displacement vs time curves for the AR and TMT conditions, and the representation of Ln

Figure 8 .
Figure 8. left side images: tEM micrographs showing the presence of MX nanoprecipitates pinpointed by white and black arrows.the 1 at.%V isoconcentration surface (right side reconstruction), and the 1 D concentration profile (middle plots) of elements Cr, V, W, Co, C and n, performed along the nanoprecipitate denoted by the black arrow in the isoconcentration surface for the steels: (a) HDsn1, (b) HDsn2 and(c)n HDsn3 after the tempering treatment.the tEM images and the apt results are not correlated but do refer to the same heat treatment condition.after(Vivas et al., 2020;Vivas et al., 2021).

Table 2 .
number density and mean size of M 23 C 6 carbides and MX nanoprecipitates observed in the HDsn steels.