Comment on "Liquid-Liquid Phase Transition in Supercooled Yttria-Alumina"

We welcome efforts made by Barnes et al to study liquid-liquid transitions (LLT) in supercooled Y2O3x-Al2O3(100-x) or AYx melts reported in a recent Letter [1]. The first order LLT in AY20, which we have identified in situ under equilibrium conditions [2,3] has a well-defined temperature TLL at ambient pressure and a critical point at negative pressure. The LLT is also composition dependent, rising as x falls [3,4]. We identified this at 1788K from (1) a peak in the small angle X-ray scattering (SAXS) intensity, (2) a discontinuity in the structure factor S(Q), and (3) a polyamorphic rotor caused by periodic LLTs (Figure 1A) the flipping time and associated temperature spikes yielding the LLT density and entropy discontinuities in agreement with ex situ experiments [4]. Barnes et al used similar experiments but could not reproduce (1) or (2) in AY20 at 1788K and attributed (3) to 60K instabilities sometimes encountered during sample conditioning [1]. We consider their null results for AY20 are due to (A) large neutron beam sizes in small angle neutron scattering (SANS), and (B) doubts in AYx composition. Moreover comparison with our work [2] is obscured in [1] by reliance on apparent temperatures uncorrected for emissivity [5] and by inconsistent molar normalisation leading to flaws in modelling LLTs from our data [3]. (A) In their SANS measurements [1] the mm radius droplet was overspilled by a 2mm radius 4.5 beam. Because of the spherical liquid surface this results in total external reflection and cross fire contamination up to at least 2x0.013=0.026 1. Sub mm focussed SAXS has none of these disadvantages [3] which is why it is sensitive to the rise and fall in scatter that occurs below 0.03 [2] at the LLT (Figure 1). Importantly polyamorphic rotor action was recorded by L. Hennet during the SANS experiments with periodic 150K spikes centred at 1940K (Figure 1B). Polyamorphic rotors have large repetitive spikes thermally distinct from the oscillations illustrated in [1] which sometimes occur when molten drops contain inclusions. (B) The sample preparation method of fusing 85mg drops from weight-matched beads of separate oxides [6] is unreliable without validation of recovered samples. We fused mm radius drops by weighing material from 5g powdered batches following repeated sintering and regrinding. Composition checks post experiment confirm x<mol 1% accuracy. S(Q)s of AYx melts at 2300K are composition sensitive (Figure 1). Measurements at 11-ID-C (APS) [2] and at ID11 (ESRF) [7] are in good agreement, with the positions of the first and second peaks Q1 and Q2, and size of the principal peak S(Q1) all scaling almost linearly with x, except for values fitted from the peak maxima at 2300K from [1] FIG 1. Rather than AY20 these triangulate with the composition. We already investigated AY15 until crystallisation intervened at 1927K [2] close to the rotor temperature shown in Figure 1B. If 1940K marks the LLT for AY15 then, with our observation of 1788K for AY20 [2], these results further demonstrate that for AYx liquids TLL rises as x falls [3,4], and provide scope for future collaboration