Microstructural comparison of effects of hafnium and titanium additions in spark-plasma-sintered Fe-based oxide-dispersion strengthened alloys

Two oxide dispersion strengthened alloys: 14Cr-0.25Y 2 O 3 -0.22Hf (wt.%) and Fe-14Cr-0.25Y 2 O 3 -0.4Ti (wt.%) were fabricated by mechanical alloying and subsequently consolidated by spark plasma sintering (SPS). Electron backscatter diffraction showed grain sizes in the range 0.5 e 15 m m in both alloys. Transmission electron microscopy and scanning transmission electron microscopy showed a homoge-neous distribution of nano-oxides precipitated during SPS. Using high resolution transmission electron microscopy, energy dispersive X-ray spectroscopy and atom probe tomography, several different oxide phases were found in both alloys, but the majority of dispersoids were Y-Hf-O type in Fe-14Cr-0.25Y 2 O 3 - 0.22Hf and Y-Ti-O type in Fe-14Cr-0.25Y 2 O 3 -0.4Ti. There were a variety of orientation relationships between the different dispersoids and the ferritic matrix. Both alloys had dispersoid densities of ~10 23 /m 3 , with average diameters of 4.3 nm and 3.5 nm in the 0.22Hf and 0.4Ti containing alloys, respectively. Per atom added, Hf (0.07 at.%) is suggested to be more potent than Ti (0.46 at.%) in re ﬁ ning the nano-oxides. © 2017 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).


Introduction
Oxide dispersion strengthened (ODS) alloys have been studied intensively for a wide variety of applications [1e5] because of their excellent high temperature strength and improved creep resistance. In a radiation environment in which materials are exposed to an ion or neutron flux, the high density of oxide dispersoids are able to trap or act as a sink for vacancies, interstitials and gas atoms, improving resistance to radiation damage and to gas accumulation [6].
ODS alloys are typically fabricated by mechanical alloying of an alloy powder and a minor fraction of oxide powder, followed by consolidation through hot extrusion or hot isostatic pressing (HIP) of degassed cans [1,7e10]. Recently, spark plasma sintering (SPS), also known as the field assisted sintering technique (FAST), has been introduced as an alternative for the consolidation of powders, including ODS-based powders [11e14]. The principal advantages of SPS/FAST are to reduce the thermal exposure of powders (e.g. typically 1000e1200 C for 4 h in HIP), to reduce processing times and costs, and potentially to better retain metastable microstructural features such as fine grains of one micron or smaller.
In Fe-based ODS alloys proposed for structural applications in the nuclear fission or fusion power plant environment, titanium additions, typically to a Fe-14Cr (wt.%) ferritic type base alloy, have been shown to produce a refining effect on the resulting dispersoids, and a change from Y-O to Y-Ti-O containing oxides [15e19], although the precise mechanism for refinement has not yet been clarified e.g. the balance between promotion of nucleation or growth restriction. Nonetheless, these studies suggest the possibility that other elements may show a similar refining effect, and by understanding the potency (or not) of other elements, greater understanding on the role of Ti may be obtained, or more powerful refiners discovered.
In this paper, we build on the relatively few papers on the effect of hafnium addition to ODS alloys [20e23], for example the suggested refining effect of hafnium on oxides in austenitic ODS stainless steels [20], by using rapid consolidation by SPS/FAST of ferritic powders followed by the application of a range of highresolution characterization techniques, including transmission electron microscopy (TEM), scanning transmission electron microscopy (STEM), energy dispersive spectroscopy (EDS), and atom probe tomography (APT). To elucidate the effect of Hf, we study "back-to-back" Fe-14Cr-0.25Y 2 O 3 -0.22Hf (wt.%, "14 YH") and Fe-14Cr-0.25Y 2 O 3 -0.4Ti ("14 YT"), and the characterization focuses on a comparison of the resulting dispersoid types. Both alloys contained a high dispersoid number density of 10 23 /m 3 . In 14 YH, most dispersoids were composed of Y-Hf-O type with an average size of 4.3 nm while in 14 YT, most dispersoids were composed of Y-Ti-O type with an average size of 3.5 nm.

ODS alloy processing
Argon gas atomized Fe-14Cr wt.% pre-alloyed powders (<150 mm in diameter, Aubert & Duval, France), elemental Hf and Ti powder (À325 mesh, 99.5%, Alfa Aesar, UK) and Y 2 O 3 powder with a particle size ranging from 20 to 40 nm (Pi-Kem Ltd, UK) were used as starting materials for the mechanically alloyed powders with nominal composition Fe-14Cr-0.25Y 2 O 3 -0.22Hf (wt.%, "14 YH") and Fe-14Cr-0.25Y 2 O 3 -0.4Ti (wt%, "14 YT"). The atomic percentage of Hf in 14 YH and Ti in 14 YT was 0.07 and 0.46 at.%, respectively. Milling was performed in 99.999% purity Ar using a planetary ball mill (Pulverisette 6, Fritsch GmbH, Germany) with a chrome-steel bowl (500 ml), AISI 52100 steel balls with a ball-to-powder weight ratio of 10:1 and a rotational speed of 150 rpm. A steady-state condition/ microstructure of the 14 YH and 14 YT powder was achieved after milling for 60 h s [39]. The milled powder was loaded into a graphite mould lined with graphite paper in an Ar-filled glove box for SPS consolidation.
SPS consolidation was performed at FCT Systeme GmbH, Rauenstein, Germany, in a vacuum of 5e8 Pa at 1150 C for 5min under a peak uniaxial pressure of 50 MPa. The heating rate applied was 100 C/min. The dimensions of the consolidated 14 YH and 14 YT samples were typically 20 mm in diameter and 5 mm in thickness; further details are available in Ref. [24].

TEM specimen preparation
Alloys 14 YH and 14 YT were mechanically polished to a 4000 grit finish using SiC paper. Transmission electron microscopy (TEM) specimens were then prepared by focussed ion beam (FIB) lift-out using 30 kV and finally 5 kV Ga þ ions in a Zeiss Auriga FIB fitted with a scanning electron microscope (SEM). The Pt grids used were made by cutting a standard Pt mesh grid in half so that the gridends could be used for hosting lift-out samples (Fig. 1a). After being welded by Pt deposition to the grid, lift-out foils were further thinned by 30 kV Ga þ ions at 1.5 tilt respect to the sample surface and finally 5 kV Ga þ ions at 7 tilt respect to the sample surface, reaching a 120e150 nm sample thickness (estimated from the FIB image) with three thicker edges retained as a "frame" to hold the lamella (Fig. 1b).
Near-surface ion-beam damage on FIBed specimens is inevitable and can hamper micro-structural analysis [25e27]. Calculations using SRIM (Stopping and Range of Ions in Matter) suggested ã 20 nm deep damaged layer from the 30 kV Ga ions at 1.5 tilt respect to the sample surface and a~7 nm deep damaged layer from the 5 kV Ga ions at 7 tilt respect to the sample surface (Fig. 2). Hence, a final flash electro-polishing procedure was used to remove the damaged layer on both sides of the lamellae [28]. The specimen, held in Au coated tweezers, acted as the cathode and a curved stainless steel sheet was used as the anode. The solution was stirred during electro-polishing using a ceramicecoated stirring bar, spun magnetically. The polishing solution was 4% HClO 4 þ 96% ethanol at À30 C, with a polishing voltage of 12 V. Polishing times were typically 0.2 s, removing~20 nm from each side of the foil. After polishing, the foil was rinsed with ethanol at À30 C and then ethanol at room temperature. Specimens were transferred to TEM examination right after electro-polishing. Fig. 3 shows an example of typical TEM micrographs of Fe-14Cr-0.25Y 2 O 3 (wt.%) before and after flash electro-polishing, indicating the importance of the flash electro-polishing approach.

APT specimen preparation
Samples for APT characterization were prepared by mechanically cutting 0.5 Â 0.5 Â 15 mm rods from the bulk material. The rod sides were mechanically polished to a 1200 grit finish using SiC paper, followed by electropolishing in two stages at room temperature to produce needle-shaped samples with a tip radius <100 nm [34,35]: Stage 1: Voltage between 15 and 7 V in a 25 vol% perchloric acid (60%) and 75 vol% acetic acid solution; and.

Characterization techniques
Alloy grain structures were investigated by electron backscatter diffraction (EBSD) in a JEOL 6500F operated at 20 kV, a probe current of 10 nA, a probe step size of 50 nm and a mapping area of 8 Â 20 mm 2 . More detailed microstructural studies, including investigating the structure and interfaces of dispersoids were carried out by TEM using a JEOL 2100 at 200 kV and a JEOL 3000F at 300 kV. Chemical profiles of the dispersoids and interfaces were examined by STEM-EDS in a JEOL 2100.
Atom probe tomography (APT) analysis was carried out in a CAMECA LEAP 3000HR™ instrument operating in laser-pulsing mode at a repetition rate of 200 kHz, using a laser energy of 0.4 nJ and maintaining a base temperature of 50 K. APT data reconstruction was carried out using CAMECA IVAS™ software. Y, Y-O and Hf-O were selected as core-cluster ions. The maximum separation method was used to identify clusters [36]. A minimum number (N min ) of 9 ions in a maximum separation distance (d max ) of 1.2 nm were taken as the cluster selection parameters, finding 162 clusters in the analyzed volume. 3D chemical maps with atomic spatial resolution were reconstructed. The cluster size was calculated in terms of the radius (or diameter) of gyration [37]. Best-fit ellipsoids were applied to characterise the shape of the clusters. The smallest, middle and largest characteristic lengths of the bestfit ellipsoids allow estimation of the oblateness (smallest characteristic length/middle characteristic length) and aspect ratio (middle characteristic length/largest characteristic length) of the clusters, defining sphere, rod, lath or disc shapes [38]. Fuller details of the APT characterization of alloy 14 YT are given in Ref. [24]. Fig. 4(a) and (b) show EBSD grain orientation maps for the 14 YH and 14 YT alloys. All grains were ferritic and grain orientations (surface normal) are colour-coded as in the inset standard triangles. Grain boundaries were defined as having misorientation >5 and sub-grains as having misorientation between 0 and 5 . Both 14 YH and 14 YT had random grain orientations and grain sizes in the range 0.5 mme15 mm, with some occasional very large grains, as shown in Fig. 4a. Previous studies [29,30] have shown that inhomogeneity of the yttrium distribution associated with inhomogeneous precipitation and pinning of grain boundaries, along with abnormal grain coarsening behavior, can lead to the formation of occasional larger ferrite grains.

TEM and STEM
The TEM micrographs in Fig. 5 show that the ferrite grains of both alloy types had relatively low dislocation densities, implying that much of the cold work of room temperature high energy ball milling was eliminated, through annealing and recovery during the short time at elevated temperature during SPS. Dispersoids in 14 YH and 14 YT appear to be homogeneously distributed. Also, no porosity was observed in either alloy.
The STEM annular dark field (ADF) micrographs in Fig. 6 show details of typical dispersoids. In 14 YH (Fig. 6 a) the dispersoids were usually as small as 1e5 nm in diameter but with a few larger dispersoids of >30 nm and the dispersoids were mostly spherical with a few of cubic morphology. In 14 YT (Fig. 6 b), dispersoids were more uniform in diameter and were mostly cubic while the larger dispersoids (ranging from 9 nm to 15 nm) were mostly spherical. In  both alloys, some dispersoids (circled in white) showed misfit Moir e fringes confirming that the particles were crystalline and potentially coherent or semi-coherent (partial coherent) with the ferritic matrix [31,32].
The size of~1300 dispersoids in each material was investigated using Image J image analysis software. The average size of dispersoids was described as a diameter if they were near-spherical, as a diagonal distance if cubic or rectangular, or as the longest internal  distance if more irregular in shape. In 14 YH, the average dispersoid size was~4.3 nm with a density of 0.8 Â 10 23 /m 3 while in 14 YT the average dispersoid size was~3.5 nm with a density of 1.04 Â 10 23 / m 3 . Fig. 7 compares the size distributions obtained from 14 YH and 14 YT, with dispersoids smaller than 10 nm comprising~92% in 14 YH and~95% in 14 YT. Fig. 8(a) shows an HR-TEM micrograph of the matrix of alloy 14YH with a highlighted dispersoid while Fig. 8(b) shows the corresponding FFT image on Z ¼ [111]. Matrix planes are indexed in white while the related planes in the dispersoids are indexed in yellow. The dispersoid inter-planar spacing was 0.249 nm and was coherent with the matrix on {110}. Fig. 9(a) shows an HR-TEM micrograph of another dispersoid in 14YH and its corresponding FFT image on Z [100] is shown in Fig. 9(b). The inter-planar spacings of the two planes at 75 to each other were 0.515 nm and 0.264 nm, and the dispersoid had one plane coherent with matrix on {200}. Fig. 10(a) and (b) show an HR-TEM micrograph of alloy 14YT and the corresponding FFT images from Z [100] respectively. In this case, the dispersoid in Fig. 10a had no resolvable orientation with the matrix.

STEM-EDS
STEM-EDS was used to determine the chemical composition of typical larger (>30 nm) dispersoids in Fig. 11 for alloy 14 YH. The dispersoids are shown by bright contrast in STEM-dark field (DF) and the analyzed area is indicated by the white square. The iron, chromium, yttrium, hafnium and oxygen concentration maps show that the large dispersoids contained yttrium, hafnium and oxygen. The average atomic ratio of metallic elements in the dispersoids was approximately Y:Hf ¼ 5:2. Similarly for alloy 14 YT in Fig. 12, the larger (>30 nm) dispersoids contained yttrium, titanium and oxygen but in atomic ratio of approximately Y:Ti ¼ 1:1.

APT
The smaller dispersoids in both alloys were too small for reliable analysis by STEM-EDS. However in related work, Zhang et al. [24] had previously studied the identical alloy 14 YT by APT, and     found that the smaller dispersoids (<5 nm) had a Y:Ti atom ratio close to 1:1, and an estimated dispersoid number density from APT of 2.7 Â 10 23 /m 3 with an average Guinier diameter of 1.5 ± 0.4 nm.
Following an identical APT approach used in Ref. [24] but now applied to the previously unstudied alloy 14 YH, Fig. 13 shows an APT 3D reconstruction with distinct clustering of Y, Y-O and Hf-O ions. These clusters had a number density of 5 Â 10 23 /m 3 and a Guinier diameter of 2.9 ± 0.7 nm. The corresponding ion cluster size distribution is shown in Fig. 14. The morphology of the clusters can be approximated from the dimensions of best-fit ellipsoids, as shown in Fig. 15. By APT, the majority of clusters (83%) were spherical, 9% were disc-shaped and 8% were rod-shaped. All the clusters were Y-O rich, and almost all of them (97%) were Y-Hf-O rich. Y, Hf and O concentration in the clusters is shown as a function of cluster size in Fig. 16. The smallest clusters, i.e. <3.5 nm in size, have variable stoichiometry, but those larger than 3.5 nm showed a Y:Hf ratio ¼ 5:2, as displayed in Fig. 17, and in close agreement with the earlier STEM-EDS estimations.

Discussion
In ODS steels, yttrium oxide is normally the major dispersoid type [3,33]. Titanium has been used frequently to refine the final dispersoid size distribution [18,24]. In this study, we have investigated the effect of hafnium addition, alongside comparative effects of titanium, as a dispersoid refining additive. While there have been some limited studies on the effect of hafnium on ODS dispersion e.g. in a Co-based ODS alloy [22] and an ODS austenitic stainless steel [20], the results reported here are the first for the more widely studied Fe-14Cr alloy. The choice of 0.22 wt% Hf in the study was purposely to make it comparable to a Fe-Cr-0.25Y 2 O 3 (wt. %) alloy previously prepared and characterised [41]. In another separate submitted paper [42], the effectiveness of Hf individually in Fe-14Cr is discussed as well on how Hf promotes a fine distribution of precipitates in the alloy can be compared to the effectiveness of Y.
Both Fe-14Cr pre-alloyed powder and elemental Ti and Hf powders were used as starting materials for mechanical alloying in order to fabricate ODS steels. Due to large surface area and high   reactivity of metallic powders, oxygen was inevitably absorbed on the metallic powder particle surface [40]. Both this absorbed oxygen on the metallic powder surfaces and the oxygen contained in the smaller Y 2 O 3 powders ensured sufficient oxygen for the formation of Y-Hf oxide and Y-Ti oxide during SPS processing. Overall, the effects of Ti and Hf on the microstructure of the resulting alloys containing Y 2 O 3 -based dispersoids were rather similar: the dispersoid size distributions in alloys 14 YH and 14 YT in Fig. 7 showed most dispersoids were between 1 and 5 nm in diameter, with a peak size of~2 nm and~3 nm in the alloys 14 YT and 14 YH, respectively. The dispersoid density in both Ti and Hf containing alloys was~10 23 /m 3 . However, while most dispersoids in 14 YT were of the previously reported Y 2 Ti 2 O 7 type [24], in 14 YH both STEM-EDS data and APT analyses suggested a Y: Hf atomic ratio of 5:2.
Further, the dispersoid compositions for the Hf-containing alloy found here differed from those in previous reports: Oka et al. [20] suggested Y 2 Hf 2 O 7 in an ODS austenitic stainless steel with 0.6 wt % Hf addition; and Zhang et al. [22] suggested that in Co-based ODS alloys two kinds of Y-Hf-O complex were present depending on the Hf concentration: Y 2 Hf 7 O 17 (1Y 2 O 3 :7HfO 2 ) at 2.4 wt% Hf and Y 2 Hf 2 O 7 (1Y 2 O 3 :2HfO 2 ) at 1.2 wt% Hf. Neither of these provides a close match to the ratios suggested by both STEM and APT in the present study. There was no strong evidence for Y 2 O 3 or HfO 2 i.e. so far as was resolvable using the techniques employed here, Y and Hf always interacted with O together. This difference may arise from the lower Hf concentration used here, which resulted in a further Hf(þY) containing oxide variant, or subtle differences in the processing history and unresolved differences in the tramp impurity content. Nonetheless, a key finding from the present work is that atom for atom, Hf provides a stronger refining effect on the dispersoid size than Ti.

Conclusions
A variety of high resolution characterization techniques have been used to investigate two ODS alloys fabricated by MA followed by SPS consolidation: Fe-14Cr-0.25Y 2 O 3 -0.22 Hf (14 YH) and Fe-14Cr-0.25Y 2 O 3 -0.4 Ti (14 YT). Compared with the conventional HIP route for consolidation of ODS alloys [7], where pressures of several hundred MPa and times of several hours are needed, SPS successfully consolidated the alloys at much lower pressure (50 MPa) and shorter time (5 min). The grain structures were similar to those obtained from HIP processing of the same alloys. Yttrium-rich dispersoids were distributed homogeneously in the consolidated alloys, with a number density of~10 23 /m 3 , which is typical for alloys of this type. Hf and Ti additions had similar effects on refining the dispersoid size, even at relatively low Hf concentrations (0.07 at.% Hf vs. 0.46 at.% Ti). Nanoscale chemical data suggested that Hf did not simply substitute for Ti in the dispersoids. While it is likely that both Ti and Hf play a critical role in promoting the nucleation of dispersoids, either directly by forming heterogeneous nucleation sites and extremely fine-scale transition phases, or indirectly by promoting higher dislocation densities, the precise mechanism has yet to be identified. Nonetheless, atom for atom, Hf provided a stronger refining effect on the dispersoid size than Ti.