Grain boundary segregation engineering in nanocrystalline alloys

Abstract

Solute segregation to grain boundaries can stabilise nanocrystalline alloys, but the same interfacial chemistry may also change boundary mobility, local cohesion, and deformation mode. We report a comparative experimental and thermodynamic study of segregation engineering in three mechanically alloyed nanocrystalline systems: Cu-1.2 at% Zr, Cu-0.8Zr-0.4Hf, and Fe-10Cr-0.6Zr. Atom probe tomography, transmission Kikuchi diffraction, aberration-corrected scanning transmission electron microscopy, differential scanning calorimetry, and microcompression were used to connect grain-boundary excess, boundary complexion state, thermal stability, and small-volume plasticity. The Cu-Zr-Hf alloy retained a median grain size of 43 +/- 7 nm after annealing for 100 h at 0.58 Tm, compared with 96 +/- 18 nm for binary Cu-Zr and 121 +/- 24 nm for solute-lean Cu. The Fe-Cr-Zr alloy showed weaker grain-size retention but a smaller loss of compressive strain to failure. Interfacial excess measurements indicate that submonolayer Zr coverage lowers the apparent grain-growth mobility by approximately one order of magnitude, whereas Hf co-segregation promotes discontinuous amorphous complexions at selected high-angle boundaries. Those complexions further suppress grain growth but concentrate shear during microcompression. The results support a design window in which segregation should be sufficient to reduce boundary energy and mobility, but not so strong that thick amorphous films percolate through the boundary network.

Introduction

Nanocrystalline metals combine high interface density with short diffusion distances and can exhibit strength levels that are inaccessible in conventional coarse-grained alloys [1]. Their practical use is limited by grain growth during processing, joining, and service. Because the driving force for grain growth scales with boundary area, a nanocrystalline alloy with 20-50 nm grains can be thermally metastable even at temperatures where the bulk phases remain unchanged. Stabilising this structure therefore requires control of the boundary network itself, not only refinement of the starting powder or film.

Segregation engineering provides one route to that control. Thermodynamic models show that solutes with favourable grain-boundary segregation enthalpy can reduce the excess free energy of the boundary network, while kinetic effects such as solute drag and second-phase pinning can further reduce grain-boundary mobility [2,3,4,5]. The idea is attractive because it treats grain boundaries as alloy-design targets rather than unavoidable defects. In practice, however, boundary chemistry is heterogeneous. Segregation depends on grain-boundary character, local curvature, competing solutes, excess free volume, and the non-equilibrium state introduced by severe plastic deformation or mechanical alloying [6,7,8,9].

Recent work has also shown that segregation does not merely decorate otherwise unchanged boundaries. Sufficient solute enrichment can produce complexion transitions, including nanoscale amorphous films that alter mobility and mechanical response [10,11,12]. Such complexions may be beneficial for thermal stability, but continuous weak boundary films can reduce strain hardening or promote localised shear. The central question is therefore not whether more segregation is always better; it is how much segregation, on which boundaries, and in which chemical form produces a useful balance of stability and damage tolerance. The present study addresses this question using three model alloys chosen to separate submonolayer segregation, co-segregation, and complexion-mediated stabilisation.

Materials and experimental procedure

Elemental Cu, Fe, Cr, Zr, and Hf powders of 99.9% nominal purity were blended to target compositions of Cu-1.2 at% Zr, Cu-0.8 at% Zr-0.4 at% Hf, and Fe-10 at% Cr-0.6 at% Zr. A solute-lean nanocrystalline Cu reference was prepared under the same milling schedule. Powders were cryomilled for 14 h under argon using hardened steel media and a 10:1 ball-to-powder mass ratio. Oxygen pickup was monitored by inert-gas fusion and remained below 0.18 wt% in the copper alloys and 0.11 wt% in the iron alloy. The milled powders were sealed under argon, compacted at 450 MPa, and consolidated by spark plasma sintering for 6 min at temperatures selected to avoid measurable bulk precipitation in preliminary DSC scans.

Discs 12 mm in diameter and 2 mm thick were sectioned into coupons for annealing at homologous temperatures between 0.46 and 0.62 Tm. Isothermal holds were performed for 1, 10, and 100 h in evacuated quartz ampoules backfilled with 5 mbar argon. Grain size was measured by transmission Kikuchi diffraction using a 20 nm step size for as-consolidated material and a 35 nm step size for annealed material. The reported median equivalent-circle diameters were computed after removing grains with fewer than five indexed pixels and correcting for the two-dimensional section bias using the Saltykov method.

Atom probe tomography specimens were prepared by focused ion beam lift-out from both as-consolidated and annealed coupons. Reconstruction parameters were adjusted using crystallographic poles where present and were held fixed within each alloy condition. Grain-boundary excess was extracted by proximity-histogram integration across manually identified interfaces, with the matrix composition estimated from adjacent grain interiors. Because trajectory aberrations and local magnification can bias boundary compositions, the absolute values are treated as semi-quantitative; trends between alloys and annealing conditions are more robust. STEM-EDS was used as an independent check for Zr- and Hf-rich boundary films thicker than roughly 1 nm.

Mechanical response was assessed by room-temperature nanoindentation and by compression of focused-ion-beam milled pillars with diameters of 900 +/- 80 nm and aspect ratios between 2.2 and 2.6. The pillar geometry is not intended to provide a bulk tensile ductility measurement. It was used instead to compare strain localisation tendencies between boundary states after normalising for grain size and residual porosity. At least 18 pillars were tested per condition using a constant engineering strain rate of 10^-3 s^-1.

Segregation and grain-growth analysis

The design analysis followed the thermodynamic stabilisation framework proposed for nanocrystalline alloys [2,4]. The total free-energy change was written as the sum of chemical mixing, elastic strain, and interfacial terms. Solute segregation reduces the interfacial term when the boundary excess lowers the boundary free energy more than the bulk mixing penalty. We did not treat the model as a quantitative predictor of final grain size, because real mechanically alloyed powders contain stored dislocation density, oxide fragments, and local composition fluctuations. Instead, the model was used to define expected trends in segregation strength and to select annealing temperatures where differences in mobility could be measured within practical hold times.

Grain growth was fitted using d^n - d0^n = kt, where d is the median grain diameter and n is an effective growth exponent. A single exponent was fitted over all temperatures for each alloy, while k followed an Arrhenius form. This approach is deliberately coarse: it collapses a distribution of boundary characters and mobilities into a scalar mobility. The fit is nevertheless useful for comparing stabilisation mechanisms. If thermal stability were controlled only by a reduced driving force, one would expect the apparent activation energy to change modestly. If solute drag or complexion drag dominates, both the effective exponent and activation energy can shift.

To connect chemistry with mobility, we estimated the Gibbsian interfacial excess Gamma for Zr and Hf at 63 boundaries in the copper alloys and 41 boundaries in Fe-Cr-Zr. Boundaries were grouped by misorientation angle and by whether APT or STEM showed evidence of a film-like complexion. The statistical power is limited, particularly for low-angle boundaries, but the distribution is more informative than a single mean excess. Prior work has shown that segregation energies in polycrystals are broadly distributed rather than unique material constants [8], and our analysis follows that view.

Results

The as-consolidated median grain sizes were 31 +/- 5 nm for solute-lean Cu, 36 +/- 6 nm for Cu-Zr, 34 +/- 6 nm for Cu-Zr-Hf, and 48 +/- 9 nm for Fe-Cr-Zr. After 100 h at 0.58 Tm, the solute-lean Cu reference coarsened to 214 +/- 42 nm, while Cu-Zr reached 96 +/- 18 nm and Cu-Zr-Hf reached 43 +/- 7 nm. The Fe-Cr-Zr alloy was tested at a lower absolute temperature but comparable homologous temperature; it coarsened from 48 +/- 9 nm to 112 +/- 21 nm after 100 h at 0.56 Tm. These values place the ternary Cu-Zr-Hf alloy in the strongest stability class of the tested set, but not in a completely pinned state. Grain growth remained measurable at the longest hold time.

APT revealed Zr segregation at most high-angle boundaries in Cu-Zr, with a median excess of 1.6 atoms nm^-2 after annealing. In Cu-Zr-Hf, Zr excess decreased slightly to 1.3 atoms nm^-2, but Hf contributed an additional median excess of 0.8 atoms nm^-2. The combined distribution was strongly skewed: roughly 18% of analysed boundaries had total Zr+Hf excess greater than 3 atoms nm^-2. STEM-EDS showed discontinuous 1-2 nm solute-rich films at a subset of those boundaries, consistent with complexion formation. No continuous amorphous boundary network was observed in the sampled volumes. The Fe-Cr-Zr alloy showed weaker Zr segregation, with a median excess of 0.7 atoms nm^-2, and Cr enrichment at boundaries was boundary-specific rather than universal.

The effective grain-growth exponent was n = 3.1 +/- 0.4 for solute-lean Cu, 4.8 +/- 0.6 for Cu-Zr, 6.2 +/- 0.7 for Cu-Zr-Hf, and 4.0 +/- 0.5 for Fe-Cr-Zr. Apparent activation energies increased from 82 +/- 9 kJ mol^-1 for solute-lean Cu to 146 +/- 15 kJ mol^-1 for Cu-Zr and 203 +/- 21 kJ mol^-1 for Cu-Zr-Hf. These fitted quantities should not be over-interpreted as elementary diffusion barriers. They indicate that the boundary network in the ternary alloy evolves through a slower, more heterogeneous mobility distribution than the solute-lean reference.

Room-temperature hardness scaled approximately with grain size across the annealing series, but the microcompression response was sensitive to boundary chemistry at fixed grain size. As-consolidated Cu-Zr-Hf pillars reached 2.3 +/- 0.2 GPa flow stress at 2% strain, compared with 2.0 +/- 0.2 GPa for Cu-Zr. However, Cu-Zr-Hf showed abrupt load drops in 39% of tested pillars after annealing at 0.58 Tm, whereas Cu-Zr showed load drops in 17%. Post-mortem SEM indicated shear offsets that intersected solute-rich boundaries in several Cu-Zr-Hf pillars. The Fe-Cr-Zr pillars had lower flow stress, 1.7 +/- 0.2 GPa, but a broader distribution of stable plastic strain before localisation.

Discussion

The results support a two-regime picture of segregation engineering. At moderate boundary excess, Zr segregation reduces the thermodynamic driving force for grain growth and produces solute drag without fundamentally changing boundary structure. This regime corresponds to the Cu-Zr alloy and is consistent with thermodynamic stabilisation models and previous experimental work on nanocrystalline alloys [2,4,5]. The grain-growth reduction is substantial but incomplete, which is expected because not all boundaries offer the same segregation energy or mobility.

The Cu-Zr-Hf alloy enters a second regime in which co-segregation promotes discontinuous amorphous complexions. The stronger grain-size retention agrees qualitatively with reports that complexion formation can enable exceptional thermal stability in nanocrystalline alloys [10,13]. The mechanical data, however, show that this stability is not free. Boundaries with film-like excess appear to participate in strain localisation during microcompression. This does not mean that amorphous complexions are intrinsically harmful; they may improve boundary sliding resistance, irradiation tolerance, or high-temperature stability under other conditions. It does mean that a stability metric based only on retained grain size can select boundary states that are not optimal for room-temperature load bearing.

The Fe-Cr-Zr results are a useful counterexample. The alloy was less stable than Cu-Zr-Hf even though Zr is a strong segregant in many ferritic systems [14,15]. APT suggests that the Cr-rich matrix and boundary-character distribution modify the effective segregation landscape. Some boundaries were enriched in Cr but depleted in Zr, while others showed the opposite trend. Such competition is consistent with simulations and experiments showing that segregation in polycrystals is a network property rather than a single interface reaction [8,9,11]. Alloy design therefore requires a distribution-aware view: the most mobile boundaries, not the average boundary, can control coarsening.

Several limitations should be noted. The alloys were produced by mechanical alloying and spark plasma sintering, so residual oxygen, porosity, and stored strain may influence both segregation and grain growth. The APT sampling volume is small relative to the full boundary network, and low-angle boundaries are underrepresented. The microcompression tests probe small-volume plasticity and should not be read as bulk tensile ductility. Finally, the growth-law fits compress a complex distribution of boundary mobilities into one exponent and one activation energy. Future work should combine in situ annealing microscopy with boundary-resolved chemistry to test whether the most enriched boundaries are also the least mobile.

Conclusion

This study compared segregation-engineered nanocrystalline Cu-Zr, Cu-Zr-Hf, and Fe-Cr-Zr alloys using matched processing, annealing, microscopy, and micromechanical tests. Zr segregation alone reduced grain growth substantially relative to solute-lean Cu. Hf co-segregation further improved thermal stability by promoting discontinuous solute-rich complexions, but those boundary states increased the probability of strain localisation during microcompression. Fe-Cr-Zr showed intermediate stability and demonstrated the importance of competing solutes and boundary-specific segregation.

The main implication is that nanocrystalline alloy design should target an interfacial chemistry window, not simply maximise segregation. Submonolayer segregation can lower boundary energy and mobility while preserving a connected crystalline boundary network. Thick or percolating complexions may be useful for high-temperature stability but require separate validation for mechanical reliability. A practical design workflow should therefore report both retained grain size and boundary-state-sensitive mechanical metrics.

Data and code availability

Processed TKD grain maps, APT composition profiles, microcompression curves, DSC peak-fitting scripts, and grain-growth fitting notebooks are provided in the supplementary archive. Raw APT detector-event files are available from the corresponding author on reasonable request because of file size and instrument-license constraints. The MATLAB and Python scripts were tested with MATLAB R2021b and Python 3.9.