Description And Prediction For Alloy's Glass Forming Ability And Preparation Of Bulk Metallic Glasses

To quickly and reasonably predict the glass-forming ability (GFA) of the alloys is important for the exploitation of new metallic glasses. Based on the well-known bulk metallic glass systems such as Pd-based, Mg-based, Zr-based and La-based, et al, the research has been done on the relationships between critical cooling rate Rc and mixing entropy Smix, the ratio of atomic radius differenceΔd, the ratio of atomic electronegativity differenceΔe , ionic index of chemical bonding v, d / e =Δd /Δe, melting heatΔHm and free energy differenceΔG of solid and liquid respectively. And then, radial base function (RBF) models effectively used for the prediction of the GFA are established. Furthermore, the metallic glass samples are prepared by vacuum injection method and the rationality of above-mentioned relationships and RBF models is directly verified.It is shown that there are extremum phenomena between the Rc andΔd ,Δe and v respectively. However, the positions of the extremum are different from different alloy systems and the extremum phenomena may be related to the optimum composition region and the optimum additive quantity. More elements in the alloy are in favor of the improvement of the GFA. And as for the high GFA alloy, the atomic percentage, atomic radius difference and atomic electronegativity difference are harmoniously matched. The more similar the major element numbers and species in the alloy system are, the more reliable and precise is the model.A new criterion i.e. d/e criterion, for the determination of GFA is obtained by integrating theΔd and theΔe . The upward parabola relationship between d/e and Rc shows better consistency than that of theΔd's or theΔe's, which suggests that the GFA of the alloy with moderate d/e is high. Therefore, to correctly predict the GFA should not overemphasize single factor but comprehensively consider the factors. Furthermore, the second derivative value B2 of the models reflects the sensitivity of GFA to the composition and the optimum composition region of the GFA. The bigger the B2 is, the more sensitive is the GFA to the composition and the narrower is the optimum composition region.Based on the hypotheses and thermodynamic theory, a series ofΔG models for metallic glass are derived and divided into three categories, calledⅠ,ⅡandⅢwhich are suitable for different alloys respectively. Furthermore, a new parameter called limit temperature Tj (defined as the lowest temperature of the model forthe calculation of theΔG) is put forward. Interestingly, the result of theΔG is related to the melting temperature Tm, i.e., theΔG curve"shifts"with different the Tm. And then, the"shift rage"is related to the melting regionΔT, i.e., the bigger theΔT is, the bigger is the"shift rage"of theΔG curve. Therefore, a rule called mono-orientation absolute transition rule (MOATR)for the choice of theΔG model is provided, and the minimum estimation values of reduced glass transition temperature Trg and theΔT for the transition of theΔG curve are obtained. And the Rc is not linearly related to theΔG. In addition, the equivalent melting temperature rTm and the equivalent coefficientβ( 0≤β≤1)are proposed, when the Tm randΔG are moderate of the alloy, its GFA is optimum. The condition of lessΔG is at bigger Trg and moderateΔT . In addition, it is validated the rationality and limitation of the Trg for the GFA criterion by theoretical analysis and the test. The test results and theoretical analysis show that the calculated Rc by MOATR is tallied with corresponding tested values not only in the magnitude but also in the sequence of the GFA in spite of heating rates B and the Tm, which shows the calculated Rc by MOATR is less related to the B and the Tm. In addition, the calculated Rc by MOATR, formula (8-5) and Johnson model and the predicted Rc by RBF are coherent with corresponding tested values in the magnitude and the sequence of the GFA, which shows the rationality and validity of MOATR. A simple parameter, i.e.ΔHm for the determination and the description of the GFA is proposed, the predicted Rc is coherent with corresponding tested value. However, the tested results show that the predicted Rc byΔHm criterion is worse than that by the MOATR, which shows the effect of the Trg on the predicted Rc could not be neglected. Furthermore, the predicted Rc by d/e andΔHm criterion is badly coherent with corresponding tested value in the magnitude, specially worse is the latter, which mainly results from two criteria only cover part of the factors influencing the GFA of the alloy.The sutdy of RBF shows that to optimize the numbers and the species of independent input parameters and learning samples is critical for establishing RBF with better performance. Under fixed learning samples, to appropriately add samples to RBF could improve the local even whole performance of RBF and other local performances may not be influenced. To appropriately add high linear correlation input parameters in RBF could adjust inner fine structure of RBF and improve the performance of RBF. In addition, the tested results are coherent with the predicted values by RBF, which shows RBF could be used to successfully predict the GFA in order to greatly cut down the energy and capital. But RBF doesn't possess extrapolated capacity.The tests show that RBF can reasonably and reliably predict Trg,ΔTx, Zmax and Rc of the alloys. However, MOATR, d/e andΔHm can only predict Rc. In addition, the tendency of Rc predicted by four methods agrees with each other. Whereas, the values of Rc predicted by RBF and MOATR are close to those of the corresponding tested values, the values of Rc predicted by d/e andΔHm are far from those of the tested values. Especially, the Rc predicted byΔHm is the worst. Moreover, four methods can compare each other.The sensitivity of characteristic temperatures of metallic glass differs from different heating rate B. And above-mentioned sensitivity is different from different alloy systems even minor change of alloy composition. In addition, above-mentioned sensitivity is also different from the deviating the degree from the eutectic composition, namely, the closer the composition to the eutectic is, the less is the sensitivity. Furthermore, the nature logarithm lnT of characteristic temperature is linearly related to the heating rate B. The glass transition activation energy Eg and onset crystallization temperature Tx are different from different alloy system and different composition, and both are the maximum for eutectic composition. In addition, the activation energy E shows the linear relation to the sensitivity index D of the characteristic temperature to the heating rate.