Abstract
The precipitation process of the martensitic PH stainless steel Nanoflex during aging at 475°C has been investigated with atom probe tomography. The composition of the matrix and the nanosized precipitates has been determined after aging for 5 min, 4 h, 40 h and 100 h. Also, the number density of the various precipitated phases has been measured. It is shown that the precipitation process is rather complex and that different precipitate phases form in direct contact with each other. At the early stages two families of Ni rich precipitates, Ni3(Ti, Al) and Ni3(Ti, Al, Si), form on Cu rich precipitates. Later, Cr rich precipitates form on the Ni rich precipitates. Eventually the Ni3(Ti, Al, Si) precipitates dissolve and are replaced by Ni16Si7Ti6. Precipitates rich in Mo, most probably the quasicrystalline R’ phase, are also formed in direct contact with other precipitates. The identity of the precipitates is discussed further in the paper. Even though the number density of the precipitates decreases during aging, the hardness does not decrease. This can probably be explained by an increased total volume fraction of precipitates as additional phases appear during aging.
Introduction
Low carbon precipitation hardening (PH) martensitic stainless steel is known for its very good mechanical properties, owing to small intermetallic precipitates. The precipitates, which at peak hardness are nanosized and densely distributed, are crucial for the properties of the steels and a lot of effort has been put into studies of phase separation in these steels.1–4 An interesting steel of this class displaying a very strong tempering response is Nanoflex,5 which was developed by Sandvik Materials Technology. The strength of this steel ultimately reaches 3 GPa, while sustaining high ductility and fracture toughness.5,6 Aging at 475°C results in a gradual increase in strength for aging times as long as 1000 h,5 which is in contrast to other PH steels. For example, the widely used 17-4 PH starts to soften after 1–10 h of aging at this temperature.7 The unusually high strength was previously ascribed to the presence of quasicrystalline Mo rich R’ phase precipitates.5,8 However, detailed studies of the development of the steel structure during aging revealed that several precipitating phases are involved. It has been shown that small Cu clusters form together with clusters rich in Ni, Ti and Al already within 5 min at the commercially often used aging temperature of 475°C.9 The Cu clusters develop into 9R precipitates within 4 h and this phase remains at least up to 400 h.10 After aging for a few hours η-Ni3(Ti, Al) has been reported to be the dominant phase.10 Phases enriched in Mo have been observed only after long time aging (100–1000 h). Liu et al.5,8 have shown that the main Mo containing phase is the quasicrystalline R’ phase.
The precipitation process of Nanoflex at 475°C has previously been studied using the atom probe technique by Stiller and co-workers.8,9,11 In the present work atom probe tomography (APT) using a local electrode atom probe (LEAP) with a large field of view has been applied for studying the precipitation sequence at 475°C. A few results of this study were presented by Stiller et al.12 and here follows a more detailed presentation. The present study reveals a modified precipitation sequence including additional phases and it has been possible to offer an explanation to the late precipitation of the R’ phase. Also, the number density of the various phases has been measured.
Experimental
The Nanoflex steel investigated in the present work was produced and provided by Sandvik Materials Technology. Its chemical composition was Fe–13·0Cr–8·5Ni–2·3Mo–1·7Cu–1·1Ti–0·69Al–0·30Mn–0·24Si–0·03C (at-). The material was solution treated at 1050°C and water quenched. The volume fraction of martensite was maximised by cold rolling the material into strips (deformation degree of 83) that resulted in deformation induced martensite. In the subsequent aging procedure the material was aged for 5 min, 4 h, 40 h and 100 h at 475°C. The effect on hardness is shown in Table 1. A 3000X HR LEAP13 instrument from Imago Scientific Instruments was used for APT. All acquisitions were performed at a temperature of 70 K using a pulse fraction of 20 of the dc voltage and a pulse frequency of 200 kHz. The specimens were prepared by standard electropolishing methods.
Hardness development during aging at 475°C
Results
The development of the matrix composition is shown in Table 2. The values were obtained by removing all parts of the analysed volume consisting of precipitates or clusters, defined by having an iron concentration below 73. The concentration of most elements (except Fe) is generally decreasing during aging as a consequence of the precipitation process, which is described below. The evolution of the matrix composition differs in several ways from previous measurements made by one-dimensional atom probe analysis.8 For example, the decrease in Ni, Al and Ti occurring already after aging for 5 min appears to have been missed in the previous work.
Matrix composition (at-) measured with APT: errors given 2σ originate from counting statistics and minor peak overlaps
In the unaged material all elements are randomly distributed, which is in agreement with previous work.8 This was confirmed by applying various statistical tools to the APT analysis, such as nearest neighbor distance distributions and concentration frequency diagrams.14
Results illustrating the precipitation process are presented in Fig. 1. The subvolumes shown in Fig. 1 are 25×35×8 nm, so the atom maps give a two-dimensional impression. The subvolumes were chosen so as to show typical features. The entire data sets are much larger containing a few hundred precipitates for the shorter aging times.
Atom maps illustrating precipitation process: box size is 25×35×8 nm
In accordance with previous investigations,8,9 clusters of Cu have formed uniformly throughout the matrix after 5 min of aging (see Fig. 1). Beside each Cu cluster there is a relatively strong enrichment of Ni, Ti and Al and also some enrichment of Si. The co-clusters of Cu clusters and Ni rich clusters appear to be spherical with a typical radius of 1 nm. The composition of the Ni rich clusters/precipitates based on individual Ti+Al(+Si) isoconcentration surfaces is given in Table 3.
Development of Ni rich precipitates*
*For the shortest aging time of 5 min the precipitates are treated as one type. The concentrations are average values from 10 precipitates of each type and the error bars represent the standard deviation in the 10 measurements.
After aging for 4 h the former Cu clusters and the Ni rich clusters have developed into easily discernable precipitates (see Fig. 1). The two types of precipitates are always located adjacent to each other. The Ni rich precipitates can now clearly be divided into two families, one containing mainly elongated precipitates with a low Si content (∼0·3) and the other containing spherical precipitates with a high Si content (∼3) (see Table 3). It was observed that Cr and Mo were often slightly enriched at the interface between the Ni rich precipitates and the matrix. This is shown by proxigrams15,16 constructed using isoconcentration surfaces of Cu+Ni+Al+Ti and Cu+Ni+Al+Ti+Si for the Si poor and the Si rich precipitates respectively (see Fig. 2). Cu was included in the isoconcentration threshold to show enrichment at the interface between the co-precipitates and the matrix.
Proxigrams showing concentration profiles of Cr, Mo and Ti across interface between matrix and Ni rich and Cu rich 9R co-precipitates: note that maximum Mo enrichment is located closer to interface than maximum Cr enrichment at η precipitates
The Cu rich precipitates as well as the precipitates of both Ni rich families have coarsened after aging for 40 h, so that the number density is roughly one-third compared with the material aged for 4 h. The composition of the Ni rich precipitates is nearly unchanged, apart from a decrease in Fe. Furthermore, chromium is rejected from both types of Ni rich precipitates, resulting in small precipitates (typical radius of 2 nm) rich in Cr (∼56) and Fe (∼37) and slightly enriched in Mo (∼4) in direct contact with Ni rich precipitates (see Fig. 1). In principle each Cu/Ni co-precipitate is associated with one Cr rich precipitate. On the other hand, the Mo concentration in the Ni rich precipitates is similar to that in the matrix (at least regarding the Ni rich precipitates low in Si). Hence, it appears as if Mo segregates to the precipitate/matrix interfaces. This is illustrated by the proxigrams in Fig. 2.
After aging for 100 h the Cu precipitates and the elongated Ni rich precipitates have continued to coarsen. The measured composition of the Cu precipitates is 75Cu–11Fe–9Ni. The composition of the small spherical Ni rich precipitates has dramatically changed compared to the composition after 4 and 40 h. The Si content has increased from about 5 to about 16 and the Al content has decreased from about 9 to about 2 (see Table 3). In addition, the content of carbon has increased from ∼0·1 to 0·5 and Cu has decreased from 4 to 0·5. It is believed that the precipitates have transformed into G phase, Ni16Si7Ti6 (see also the discussion below). Furthermore, a phase enriched in Mo with the major composition 35Fe–34Mo–13Cr–10Ni–7Si has appeared. These precipitates vary much in size with the largest exceeding 10 nm. They are often cut by the analysis volume, which makes it difficult to determine their neighbouring environment. However, when the precipitates were entirely confined within the analysis volume they were always in contact with both a Ni rich precipitate low in Si and a Cr rich precipitate. When studying the mass balance between the composition of bulk, matrix and precipitates, it becomes clear that there is not enough Si in the material to match the decrease in Mo in the matrix via precipitation of the Mo rich phase only (also considering the Si needed for the G phase). This can be explained by Mo segregating to precipitate interfaces, as shown in Fig. 3.
Atom probe tomography reconstruction of material aged for 100 h: in centre of volume there is η precipitate outlined by Ti (blue/small) and Al (orange/small); to left is Cu rich 9R precipitate outlined by Cu (red/small) and to right is Cr rich α′ outlined by Cr isoconcentration surface (brown); it is clear that Mo (purple/large) is segregated to η/α′ interface; box size is 12×12×7 nm3, which is subvolume of analysis. The colour version of this image can be viewed in the online issue
The number density of the precipitating phases is presented in Table 4. It is clear that the number density decreases during aging, which is equivalent to saying that coarsening is taking place.
Number density (1024 m−3) of various precipitated phases
Discussion
Identification of precipitate phases
Cu rich precipitates
Several studies using electron diffraction have concluded that after sufficient aging the Cu rich precipitates are the 9R phase.10,17 At early stages the precipitates are probably coherent bcc, gradually transforming to 9R with a twinned close packed structure and possibly to 3R, or even fcc, after long enough aging.18,19 The Cu rich precipitates are distributed evenly throughout the matrix, so the nucleation appears to be more or less homogenous. In APT reconstructions the Cu rich precipitates are somewhat demagnified because the evaporation field of the precipitate is lower than the evaporation field of the matrix. This results in trajectory effects and a contribution from the surrounding matrix to the precipitate composition. Therefore, the true Cu content is likely somewhat higher than the measured 75 after 100 h of aging.
Ni rich precipitates
Already after 5 min there are clusters consisting of Ni, Ti, Al and Si adjacent to the Cu clusters. At this stage there is no clear separation between Si rich and Si poor types, but after 4 h of aging two different Ni rich phases are observed (Fig. 1 and Table 3). A slight majority of these precipitates are spherical and have a composition close to Ni3(Ti, Al, Si) with a relatively high Si content (∼3). At the same time precipitates with a composition close to Ni3(Ti, Al), practically free from Si, are observed. These precipitates are, after 40 h of aging, the largest in the material and they also comprise the largest volume fraction. The observed morphology and stoichiometry of these precipitates are in good agreement with the electron diffraction work by Wang et al., reporting about the occurrence of a hexagonal η-Ni3(Ti, Al) phase with needle-like morphology in Nanoflex after 40 h of aging.10,17 This is also the most commonly observed Ni/Ti containing phase in maraging steels and it has a [111]m//[11–20] η and (111)m//(0001) η orientation relationship with the matrix. Atom probe tomography analyses show that the precipitates have an orientation relationship with the matrix and that the angle between the orientations of the needles is ∼70°, which agrees with transmission electron microscopy (TEM) observations. This is a strong indication that the elongated Ni3(Ti, Al) precipitates are η phase. The η-Ni3(Ti, Al) phase seems to be stable, and the precipitates steadily grows and have gradually increasing Ni content at the expense of Fe and Cr at least until 100 h of aging.
The occurrence of silicon containing Ni3(Ti, Al, Si) phase in Nanoflex was not reported in any TEM study and therefore its identification is unsure. A hypothesis is that it is L12 γ′ phase. This is supported by calculations by Xu et al.20 showing that the driving force for the ordered L12 γ′ phase is large and similar to that of D024 η phase for Ni3Ti (as well as for Fe3Ti). The fact that the L12 phase also exists as Ni3Si and (Ni, Fe)3Si21,22 gives further support for this hypothesis. The phase is, however, not stable in the alloy and after 100 h of aging the presumed γ′ phase is no longer present. Instead there are small precipitates with a composition close to G phase Ni16Si7Ti6. The G phase is often observed in silicon containing ferritic steel and according to atom probe analyses C and P are enriched in this phase.23 This is also the case in the present study (see Table 3). As long as the precipitates were γ′ the level of C was much lower. Furthermore, the G phase is known to be present together with η-Ni3Ti phase in Cr/Si containing maraging steels.24–26 It also occurs in contact with Cu rich precipitates27 and its formation is promoted by the presence of Mo and Ni,23,28 which agrees well with the observations in the present work. It is therefore concluded that the precipitates are G phase.
As the γ′ transforms into G phase, Al is rejected and the Si content increases dramatically. Sometimes Al appears to be enriched at the interface between the G phase precipitates and the adjacent Cu rich 9R precipitate (see Fig. 1).
The reason why G phase is not formed earlier, for example, after 40 h, is probably that the phase cannot contain a significant amount of Al. After long enough aging, though, the η phase can consume sufficient amounts of Al. It is also interesting to note the significant drop in the matrix concentration of Ti between 40 h (0·036) and 100 h (0·009), surely related to the formation of G phase. Nucleation of G phase has been reported to often occur at α/α′ interfaces.23 In principle, this could be the case also here, which would then explain the delayed formation of G phase. However, because they are always in contact with 9R precipitates, it seems more likely that the G phase precipitates are transformed from γ′ rather than being newly nucleated at α/α′ interfaces. Also the observed enrichment of Al next to the precipitates would seem unlikely if the G phase had nucleated at α/α′ interfaces.
Cr rich precipitates
Precipitation of a high number of small spherical Cr rich precipitates was observed after 40 h of aging. There are Cr enriched areas already after 4 h, but they are regarded as enrichments at the interfaces between Ni rich precipitates and the matrix (these regions are also enriched in Mo). Previous investigations reported about σ-CrFe phase after 400 h of aging.29 The present study shows that the formation of Cr rich phase takes place much earlier. All Cr rich precipitates were in contact with Ni rich precipitates. Some of the precipitates, in materials aged for 40 and 100 h, were larger and elongated. These precipitates were exclusively in contact with elongated η-Ni3(Ti, Al) precipitates (see Figs. 1 and 3).
Two Cr rich phases could be considered as possible candidates for the observed Cr rich regions: σ and α′. The α′ phase is coherent with the martensitic matrix and can be formed in the Fe–Cr system either by classical nucleation and growth or by spinodal decomposition.30 Here the nucleation is heterogeneous so the precipitate mechanism in question must be classical nucleation and growth. Based on the presented proxigram of η and 9R after 40 h of aging in Fig. 2, it is likely that the Cr rich regions are formed due to the rejection of this element from Ni rich precipitates. One of the arguments in favour of the observed Cr rich regions being α′ and not σ is that α′ does not generate any extra diffraction patterns due to its coherence with the matrix. This could explain why this phase, despite its high number density, has not been observed in the previous electron diffraction investigations of Nanoflex.6,9,17,31 Moreover, a very strong correlation between formation of α′ phase and the presence of G phase has previously been reported23,26,27,32,33 and the work by Solomon and Levinson34 also showed that Ni and Cu addition to Fe–Cr alloys promotes decomposition.
Mo rich precipitates
Enrichment of Mo at the interface between large η-Ni3(Ti, Al) precipitates and the martensitic matrix was observed after 40 h of aging, while there was no sign of enrichment of this element at the interface of γ′ precipitates (Fig. 2). At this stage, Mo was clearly integrated into the Cr rich (α′) phase. After 100 h of aging large Mo rich precipitates appeared in contact with the co-precipitates consisting of η, α′ and 9R.
The composition measured with APT is similar to that of R′ phase analysed with TEM/EDX after 1000 h of aging (40Fe–34Mo–17Cr–2Ni–7Si)35 apart from the Ni concentration. Based on this result, in combination with the previous electron diffraction proof of the occurrence of an icosahedral quasicrystalline R′ phase,35 it is concluded that the R′ phase is present after aging times of 100 h or longer. This statement is also strengthened by the fact that another icosahedral quasicrystalline phase (I phase), which is related to the R′ phase, has been observed in the bcc matrix of other steels.36,37 In previous atom probe investigations small Mo containing precipitates with a relatively low amount of Mo were observed after 1–40 h of aging.8,11 From the present work it seems more appropriate to describe the Mo enrichment observed at η phase interfaces as segregation or in the case of aging for 40 h also as α′ precipitates.
It appears as if the precipitation of R′ does not start until after 40 h of aging and that when it eventually begins the growth is quick. It is not clear what is holding the precipitation back. A hypothesis is that precipitation starts when γ′ starts to transform to G. Even though the G phase is very rich in Si, the volume fraction is so low that the transformation and dissolution of the γ′ phase results in a supply of Si to the matrix and so Si becomes available for R′ precipitation. From the composition of R′ it can be seen that Si is the only element not available in the matrix after 4 or 40 h of aging. The low Ni content of R′ observed after 1000 h indicates that Ni is not necessary for R′ to be stable. Hence, the R′ phase can be described as a Mo–Fe–Cr–Si phase. The composition of the matrix does not change very much after 4 h of aging, so it is unlikely that it is changes in the matrix composition that suddenly make R′ favourable.
Precipitation process
Here follows an attempt to summarise the evolution of precipitation in Nanoflex at 475°C. First Cu rich precipitates (9R) form homogeneously with a high number density. More or less simultaneously, precipitates rich in Ni, Ti, Al and Si form on the Cu rich precipitates. After a short time these become either η (low Si) or γ′ (high Si). The co-precipitates start to coarsen before 4 h of aging and their number density decreases. The number density of 9R is the same as that of η and γ′ together, both after 4 and 40 h of aging (and after 100 h it is the same as that of η and G together). Judging from the matrix composition, the volume fraction of 9R is constant when going from 4 to 100 h and the volume fraction of γ′+η is rather constant when going from 4 to 40 h. The growth rate is probably suppressed by the fact that there are two families of Ni rich precipitates. Also, enrichment of Cr and Mo at precipitate/matrix interfaces probably suppresses the growth rate. After 40 h the Cr rich α′ phase is present with a number density similar to that of 9R, which agrees with the observation that there is one α′ precipitate per 9R (or Ni rich) precipitate. When going from 4 to 40 h of aging the total volume fraction of precipitates increases mainly because of the α′, as concluded from the matrix composition. When the aging has reached 100 h γ′ has transformed into G with a lower number density but with a similar size. In this process Si has been accumulated in G whereas Al has moved to η. The total amount of Si in the G phase is smaller than the total amount present in γ′. It seems reasonable that the Si released by the dissolution of γ′ is the occurrence triggering the R′ precipitation. The R′ precipitates have a low number density and are comparatively large.
Although the purpose of this paper is mainly to describe the precipitation sequence, it is worthwhile to make some attempts to estimate the hardness from the precipitation parameters. Generally, PH is considered to result from dislocation cutting when precipitates are small, and that this mechanism is gradually replaced by dislocation looping when the precipitates become large. For precipitation cutting several mechanisms have been suggested to control the strength increase, such as coherency, order and modulus strengthening.38,39 It is difficult to determine which mechanism is most relevant, in particular when there are different precipitate phases in contact with each other and when the precipitates are not spherical. However, it is often found that the hardness is proportional to (rf)1/2, where f is the volume fraction of precipitates and r is the (average) radius. This relation fits reasonably well with the parameters of the material aged for 5 min and 4 h respectively. For the extended aging times the hardness increase flattens out and the model is no longer suitable. This could indicate that dislocation looping starts to become important. However, the Ashby–Orowan equation, with the precipitation parameters of the co-precipitates, gives hardness increase values that are much higher than the observed once. For the material aged for 40 h, which has a total volume fraction of precipitates about 13, a hardness value increase of ∼830 HV is obtained. This shows that much more sophisticated models are needed for modeling the hardness increase, which is outside the scope of this work.
Conclusions
Two sets of Ni rich phases, η-Ni3(Ti, Al) and γ′-Ni3(Ti, Al, Si) are present in the matrix within 4 h of aging and all of them are in contact with Cu rich 9R precipitates. The γ′-Ni3(Ti, Al, Si) is not stable and transforms into G-Ni16Si7Ti6 during aging. Cr and Mo are enriched at the interface of Ni rich precipitates. These enrichments develop into α′ regions within 40 h of aging. Mo rich quasicrystalline R′ precipitates are present after 100 h of aging. It is proposed that R′ forms when Si becomes available following the dissolution of γ′. The precipitation process in the martensitic matrix can be summarised as
Cu rich clusters+Ni/Ti/Al/Si rich clusters (5 min)→
9R+η-Ni3(Ti, Al)+γ′-Ni3(Ti, Al, Si) (4 h)→
9R+η-Ni3(Ti, Al)+γ′-Ni3(Ti, Al, Si)+α′ (40 h)→
9R+η-Ni3(Ti, Al)+G-Ni16Si7Ti6+α′+R′ (100 h)
Despite the decrease in precipitate number density, the hardness does not decrease. This is partly explained by a gradual increase in total precipitate volume fraction, following the precipitation of new phases.
Footnotes
Acknowledgements
This work was supported by the Swedish Research Council and Sandvik Materials Technology.