Effect of Process Control Agents on Fe-15at.%Nb powder during Mechanical Alloying

06, 2021; Accepted: December 07, 2021 In this paper, Fe-15at.%Nb alloys were produced from high purity Fe (min. 99.8%) and Nb (min. 99.8%) powders via a mechanical alloying process. The effects of different Process Control Agents (i.e., methanol, hexane, and stearic acid) were investigated with powder morphologies, particle size distribution, and phase formation, and were sampled after up to 80 milling hours at 350 rpm. The powder morphologies and particle sizes were evaluated using scanning electron microscopy and laser diffraction analysis, respectively, and phases were identified via X-ray powder diffractometry. The results demonstrate for all conditions that, in the early stages, there was significant particle agglomeration due to the ductile-ductile feature of Fe and Nb powders, and latter an amorphization trend up to 80 milling hours. Methanol was the most efficient Process Control Agent in terms of avoiding cold welding, reducing of agglomeration, particle size distribution,


Introduction
Mechanical Alloying (MA) processes are characterized by repeated cold welding and powder particles fragmentation in high-energy ball mills. This welding-fracturing process eases diffusion and allows for alloy formations.
Studying Fe-Nb binary system alloys via the MA process is important because it can provide alternative processes, like viable results for better understanding the system phase equilibrium, and for many alloy families based on Fe-Nb 8,9 . Some works have detailed this [10][11][12][13][14][15][16][17][18][19] . Most studies on Fe-Nb systems 10,12,13,[15][16][17][18] deal with systemically studying phase amorphization modes via the MA process, using milling times varying from 30 minutes to 70 hours. Povstugar et al. 13 , in addition to using MA, developed a heat treatment at 700 °C. However, MA and heat treatment times were not specified for any phase that had been obtained. Vélez et al. [15][16][17] published three works related to obtaining Fe-Nb alloys via MA, based on magnetic property studies that used two heat treatments seeking LAVES (Fe 2 Nb) phase formation. Roy et al. 14 , analyzed the magnetic behavior of the LAVES amorphous phase prepared via MA. They found that the phase had paramagnetic characteristics. Blázquez et al. 19 , developed a comparative analysis of the necessary milling time for Fe-30at.%Nb and Fe-30at.%Zr alloy amorphization and the respective magnetic properties. They stated that more investigations would be needed to determine the complete Fe-Nb system phase diagram.
PCA introduced during MA has been well established a mechanism for preventing excessive cold welding, due to reduced cleaving surface energy, leading to more effective particle size reductions 20,21 . Besides, PCA has been reported preventing excessive friction between spheres and jars, mitigating wear consequentially decreasing contamination from milling tools 21 which is desirable 22,23 especially to steel industry 24,25 . As far as we know, there are no studies in literature on the PCA effect on MA in Fe-Nb system alloys. Here, the effects of different PCAs (methanol, hexane, stearic acid) on mechanical alloying for Fe-15at.%Nb alloy were studied, in terms of phase formation, particle size, distribution, and shape.

Fe-Nb System & Fe-15% at. Nb Alloy
The most recent Fe-Nb equilibrium phase diagram ( Figure 1) proposed by Liu et al. 26 has two stable intermetallic phases: (1) Fe 2 Nb (LAVES) phase, which has congruent formation (1646 °C), with C14 structures and solubility ranging between 25 (1373 °C) and 33 (1523 °C) at.%Nb; and (2) Fe 7 Nb 6 (µ) phase with peritectic formation (1523 °C), D8 5 structures and solubility ranging between 41 (1523 °C) and 51 (1508 °C) at.%Nb. The system also has two eutectic *e-mail: marcelalamoglia@unifei.edu.br reactions, one between phases (δFe) and Fe 2 Nb, and the other between phases Fe 7 Nb 6 and (Nb). Table 1 summarizes the  crystallographic data for the stable solid phases in the Fe-Nb  system 27,28 . Fe-15at.%Nb alloy composition was selected in this study because of its importance for Fe alloy industries and for new alloys that are based on Fe 26,27 .

PCA Effect
If the MA process is not controlled, excessive cold welding and high ductile particles agglomeration can occur. It is possible to introduce organic material to modify the powder surface. These are called surfactants or Process Control Agents (PCA), like methanol 20,29,30 , hexane 31 , stearic acid 20,30,31 , oxalic acid, boric acid, alumina, ethanol, polyethylene glycol, etc., added to balance the reactions 32 . These are added at 1% and 5% of the total mix, because the type and amount of PCA influences the size and shape of thin particles during milling 7,33 .
PCA molecules are based on hydrocarbon chains, which have polar functional groups. They interact with metallic powder polar groups creating covalent bonds between adsorbates (PCA) and absorbing surfaces (material). As a result, molecules form a monolayer on the powder surface. Molecular chains act as barriers around the particles to protects them from agglomeration 21,34 . Figure 2 shows PCA's role in the MA process, recovering the whole powder surface. Homogeneous PCA distribution during MA is known to be influenced by the physical state of matter (solid, liquid and gas) at room temperature. Solid state PCA (Figure 3b    In general, they decompose during milling, or interact with powder to generate compounds, and these are incorporated as inclusions and/or dispersions in powder particles during milling. Hydrocarbons, which contain hydrogen and carbon, carbon and oxygen, may introduce carbon and/or oxygen in powder particles, resulting in carbide formation and oxides, which will be evenly distributed on the matrix 7,21,30 . The latter may not necessarily be harmful to alloy systems, since they can strengthen material dispersion resulting in higher resistance and hardness 7,21,33 .
The MA process carries out in a Noah-Nuoya ® NQM 0.2 L planetary ball mill with parameters according to Table 3. The capacity was 200 g together with the milling spheres, and argon gas N4.0 (99.998% purity) was introduced inside of the milling jars to avoid oxidation. The milling rotation speed was 350 rpm, the ball to powder weight ratio was 15:1, and 60 grams of powder were used for each milling process. Samples for all conditions were collected for analysis after 20, 40, 60 and 80 milling hours.

Laser diffraction granulometric analysis
Before the granulometric analysis, samples were submitted to an ultrasound bath (Alfa Mare model ALTsonic) for 5 minutes, in a water and glycerin solution, to guarantee particle dispersion. The laser diffraction granulometric test was performed using a Microtec Flex S3500 that uses software analysis based on an optical model derived from Mie's Spreading Theory. To express a single value for the measured particle, an equivalent sphere value was adopted, defined here as average diameter.

Scanning Electron Microscopy
Micrographs were obtained via a Scanning Electron Microscope in the Secondary Electron mode using a Carl Zeiss EVO ® MA15 with 20 kV acceleration for morphologies and particle size distribution analysis.

X-ray diffractometry
Phase identification was performed using X-ray powder diffractometry (XRD) with a Panalytical X'Pert Pro ® and CuKα radiation (λ = 0.154056 nm) at 40 kV and 30 mA. The analysis was performed at a 2θ angle from 30° to 90°, with a 0.02º step size, and a 2-second counting time per step range. Phases were identified by comparing peak positions with simulated diffractograms using the PowderCell Software program 35,36 , and crystallographic information from by Villars and Calvert 27,37 .  Refining was applied to reduce XRD variation, using a Smoothing function form the Origin software program (version 9.0), and the gauss method for 15 data points.
Rietveld refinement was performed using the GSAS 2 software program. The pseudo-Voigt function was chosen because it was the most useful in describing the profiles of diffraction lines in the Fe-α peak (110), which is the most important profile in the diffractogram. The crystallite sizes and lattice parameter distortion phases were determined using the Scherer method, described in [38][39][40] . This method uses the following formula to determine the crystallite size: Where FWHM is the full width at the half-maximum diffraction peak, k is the shape constant (0.9 for polycrystalline) 41 , λ is the radiation wavelength (0.15405 nm), D is the crystallite size, and θ is the Bragg angle. Laser diffraction analysis. We should note that measurement errors were minor compared to the values, so the error bars in Figure 4 are in the same scale as the symbols.

Average particle size
The initial particle average sizes were ~ 15 µm. After 80 milling hours, for all conditions, similar particle sizes (~20 µm) were obtained. In general, we observed increases in average particle size in the initial milling stages, followed by refinement. This was expected due to cold welding at the beginning of the milling stage, followed by particle fractures in later stages when static balance occurs between cold welding and fractures, and average particle size is expected to remain the same 7 . Details on the different conditions are discussed below.
The reference sample (pure) showed increased average particle sizes in the initial milling stages, 63.3±0.3 µm for 20 hours, 93.4±0.5 µm for 40 hours, and after 60 hours the particles were refined to 73.6±0.6 µm and 28.8±0.1 µm for 60 and 80 hours, respectively.
The milled sample with methanol PCA resulted in 27.7±0.3 µm at 20 hours and 32.9±0.3 µm at 40 hours. The largest average particle size was 40.8±0.3 µm at 60 hours, suggesting that cold welding occurred between 20 and 60 hours, along with slightly increased average sizes. Particle fractures started after 60 hours, stabilizing average size between 60 and 80 hours, reaching 19.8±0.2 µm at 80 hours. The few variations among average particle sizes during the analyzed periods come from using liquid methanol, allowing for more PCA absorption on the particle surfaces in the first hours of the MA, in consequence reducing cold welding effects.
The milled samples containing liquid hexane PCA showed different characteristics compared to the previous tests. Cold welding occurred between 20 and 40 hours, and average particle size was 56.1±0.3 µm at 20 hours and remained stable at 57.7±0.5 µm at 40 hours. At 60 hours, the value decreased 51.6±0.4 µm, representing initial fractures, and at 80 hours it reached 20.5±0.1 µm at the end of the process.
For the sample with stearic acid at 20 hours MA the particles were 75.8±0.4 µm size, decreasing in the following periods to 32.5±0.3 µm at 40 h, 27.2±0.1 µm at 60 h, until stabilizing around 25.8±0.1 µm at 80 h. This is, probably, due to the solid state of powder PCA that may not be well distributed at the beginning of the milling process, resulting in less efficient powder particle absorption compared to liquid PCA 21,42 . Thus, solid agents cause more cold welding at the beginning of the MA, as was seen in this sample.
Based on this analysis and considering that cold welding is a phenomenon to be avoided in MA process, methanol is the most efficient PCA (among the analyzed) for that purpose. This behavior was expected and reported in other studies due to using liquid methanol and hexane, allowing for more PCA absorption on the particle surfaces at the beginning of the MA, and consequently, mitigating cold welding. This mechanism is described in detail by 20,21,43-46 .

Particle shapes and granulometric distribution
The particle morphology analysis focused on the final milling stage, i.e., after 80 hours, because it is expected to present reduced agglomeration and superior particle homogeneity. Particle shapes and morphology were analyzed using comparative methods given by the German classification 47 . Figure 5 shows the micrographs of the alloys after 80 milling hours. Refence (without PCA), hexane and stearic acid samples showed particle shapes in flakes (Figures 5a, 5c and 5d), while the methanol sample had polygonal shapes (Figure 5b).
The reference sample had more agglomerations in size and quantity (Figure 5a), compared to the other samples. This occurs due to the PCA ductile material, which undergoes plastic deformation under compressive loads during the MA process, causing cold welding, and consequently more agglomerations (and increased average particle size, as seen at Figure 4) 33,48,49 .
Methanol PCA (Figure 5b) and hexane PCA (Figure 5c) samples showed that agglomerates were significantly lower compared to the reference sample (Figure 5a), and the sample with stearic acid (Figure 5d). Figure 6 shows the particle size distribution under all conditions. With respect to the reference sample (Figure 6a), we can see growing asymmetric bimodal distribution, meaning population concentration composed of fine particles has been reduced. Therefore, there were more rugged particles and agglomerations. Figure 6b shows a histogram of methanol PCA, with three particle populations (trimodal), and predominant medium sized particles. By analyzing the hexane PCA samples, we noticed decreased asymmetric bimodal distribution, because of larger population concentrations if formed by fine particles, while the other population was characterized by rough particle or agglomerates (Figure 6c). The stearic acid sample (Figure 6d) also showed decreased asymmetric bimodal distribution, although its lower left peak is more scattered. As already stated, this PCA was introduced in mill in a solid state and could not dissolve nor overlay the whole surface during the 80 hours process, to counteract agglomeration.
In terms of the characteristics discussed in the present section, the methanol PCA was considered the more efficient PCA, due its wider population of particle sizes and lower degrees of agglomeration. This conclusion is based on the needs of the subsequent stages of the MA process that commonly includes pressing and particle packing that tends to be more efficient when a wider distribution of particles is present 50 .

X-ray diffractometry
From XRD diffractograms, we observed an amorphization process with increased milling times for all milling conditions. The amorphization of the microstructure is characteristic of MA processes 51,52 , and can be detected using XRD diffractograms, given the decreased intensity and enlarged peaks. This occurs because MA, which is associated with severe non-uniform plastic deformation in the lattice parameter, results in more structural defects (dislocations, vacancies, grain boundaries, etc.), decreasing the long-range order 53,54 . Crystallite average size and lattice parameter variations were also analyzed via Rietveld refining, as a function of milling time, with the Fe-α phase reference position (110). Figure 7 shows the X-ray diffractogram for the alloys without PCA. One can note the presence of Fe-α during the whole MA process, and the WC phase comes from tungsten carbide spheres. Figure 8 shows details the Fe-α (110) peak. Besides increased with increased milling times, generally associated with reduced crystal size, and increased micro-deformation, the peaks were also displaced to the left, changing for smaller angles, related to increased lattice parameters.
From Figure 9 we verified that the MA process causes quick crystallite size reductions in the initial stages. At 20 h MA this was reduced from 35.0±1.0 nm to 5.5±1.0 nm and remained in this range.
There was a progressive increase up to 40 h MA for the lattice parameter, from 2.025±0.005 Å to approximately 2.059±0.005 Å, possibly due to Nb atoms dissolving with the Fe lattice, along with accumulated defects, coming from permanent ball collisions and powder fractures, Figure 9. Figure 10 shows the X-ray diffractogram of the Fe-15at.%Nb alloy milled with methanol PCA. Besides the Fe-α phase, we identified a Nb solid solution at 20, 40 and 60 milling hours, which was not observed at 80 hours.
For the methanol PCA sample, Figure 11, shows that the peak intensity at 20 hours is well defined. As milling continues, it becomes wider and dislocates to lower angles. When the milled powder is more deformed, not only can the grain size be reduced to nanometric dimensions, but the long-range and short-range orders simultaneously decrease, leading to nanocrystalline phase formation and then, eventually, an amorphous phase 42 . Figure 12 shows gradual crystallite size reductions, from 35.0±1.0 nm, to 14.0± 1.0 nm at 20 hours, 6.0± 1.0 nm at 40 hours, and finally to 5.0± 1.0 nm at the final stage. In general, all samples with PCA tendered to gradually reduce crystallite size, with more significant reductions after 40 hours. This shows that samples required more time and energy to amorphization.
Lattice parameter values increased until 60 hours, reaching 2.045±0.005 Å. Nonetheless, this later reduced to approximately 2.036±0.005 Å. This is possibly caused by recuperation, and/or recrystallization, due to increased process temperatures 7,55 . Figure 13 shows the X-ray diffractogram for the Fe-15at.%Nb alloy milled with hexane PCA. The Fe-α and   Nb SS phases appeared after 20 hours of MA and remained in crystal form until 60 hours. At 80 hours only the most intense Fe-α peak with low crystallinity is present.
The Fe-α (110) peak showed similar behavior to samples without PCA. It increased intensity at 60 h and displaced the lattice parameter for lower angles (Figure 14).
Among the PCAs analyzed, the hexane PCA sample showed the fastest crystallinity losses in initial stages (Figure 15), from 35.0± 1.0 nm, reducing to 7.5± 1.0 nm at 20 hours and 40 hours, and finally to 4.5± 1.0 nm in final stages.
Lattice parameter values increased up to 40 hours, at 2.033 ± 0.005 Å for 20 h, and 2.055 ± 0.005 Å for 40 h. As was the case with methanol (Figure 12), there was a subtle reduction of lattice parameter values at 60 h and 80 h, 2.053 ± 0.005 Å, and 2.045 ± 0.005 Å, respectively. This is possibly caused by recuperation, and/or recrystallization, in the same sense of the sample containing methanol (Figure 12) 7,55 .
The diffractogram in Figure 16 shows the alloy milling with stearic acid PCA. As in sample without PCA, this sample showed WC contamination, justified by solid state PCA not being able to protect the whole powder surface during milling as efficiently as methanol and hexane, resulting in contamination.
Of all Fe-α (110) peaks at 20 h, the stearic acid sample had the highest intensity and lowest enlargement (Figure 17). At 40 h we verified an enlarged Fe-α peak, and at 60 h its    displaced to the left, and was even more enlargement, due to Nb atom insertion into the Fe bcc structure, increasing the lattice parameter. At 80 h the peak flattens, showing amorphization.      period, values reduce at a constant rate, with 5.0 ± 1.0 nm at 60 h and 4.0 ± 1.0 at 80 h.

Conclusions
The effects for each PCA were investigated systematically via a milling process. The morphological and granulometric analysis showed smaller average particles sizes, and smaller agglomerate formations for the material milled with PCA methanol, with particles approximately (19.81±0.21) µm, better granulometric distribution, and less cold welding.
The sample without added PCA and with solid PCA (stearic acid) showed WC contamination from tungsten spheres. Due to high friction in milling, liquid PCA (methanol and hexane) showed more efficient results because it was absorbed by the whole powder surface, protecting against contamination.
We presented the changes in crystallite size and the lattice parameter for each sample in the analyzed Fe-α peak as a function of milling time with and without PCA. We verified that the sample with methanol PCA showed the greatest crystallinity reduction rate.