Metallic Aluminum Suboxides with Ultrahigh Electrical Conductivity at High Pressure

Aluminum, as the most abundant metallic elemental content in the Earth's crust, usually exists in the form of alumina (Al2O3). However, the oxidation state of aluminum and the crystal structures of aluminum oxides in the pressure range of planetary interiors are not well established. Here, we predicted two aluminum suboxides (Al2O, AlO) and two superoxides (Al4O7, AlO3) with uncommon stoichiometries at high pressures using first-principle calculations and crystal structure prediction methods. We find that the P4/nmm Al2O becomes stable above ~765 GPa and may survive in the deep mantles or cores of giant planets such as Neptune. Interestingly, the Al2O and AlO are metallic and have electride features, in which some electrons are localized in the interstitials between atoms. We find that Al2O has an electrical conductivity one order of magnitude higher than that of iron under the same pressure-temperature conditions, which may influence the total conductivity of giant planets. Our findings enrich the high-pressure phase diagram of aluminum oxides and improve our understanding of the interior structure of giant planets.


Details of the DFT calculations
We chose 3s 2 3p 1 and 2s 2 2p 4 as the valence electrons for Al and O while using the generalized gradient approximation (GGA) in the Perdew-Burke-Ernzerhof (PBE) exchange-correlation functional [1]. The plane-wave cutoff was set as 1050 eV and the Brillouin zone (BZ) was meshed choosing the gamma-centered Monkhorst-Pack approximately 2 × 0.025 Å −1 . When calculating the electrical conductivity of iron as a comparison, the valence electrons chosen for Fe were 4s 1 3d 7  Due to the existence of the electron localization area in P4/nmm Al2O and P63/mmc AlO, we need to adjust the Wigner-Seitz radius for each atoms when calculating the electron density of states [2], which is the parameter RWIGS as implemented in VASP code. Here we set RWIGS values in accordance with the results (atomic volumes) of the Bader charge calculations rather than the parameter written in the pseudopotential.
The RWIGS for O atoms in P4/nmm Al2O is 1.05 and the RWIGS for Al atoms is 1.2.
As for the P63/mmc AlO, the RWIGS for O and Al atoms are 0.95 and 1.25, respectively.
To validate the dynamical stabilities of the predicted structures, we have investigated the phonon dispersion curves using 2 × 2 × 2 supercells with PHONOPY code [3]. Also, we have investigated the anharmonic phonon dispersion of the Cmcm Al4O7 at 1500 GPa and 300 K using the DynaPhoPy code [4]. The trajectory was provided by the Ab initio molecular dynamics simulation with a 2 × 2 × 2 supercell using VASP code.

Details of the formation energy cross-checks
We have performed cross-checks for the formation enthalpy of the Al2O using WIEN2k, VASP and CASTEP software. In the structure searches with AIRSS, the CASTEP code was employed with similar convergence parameters as used in the Vienna Ab initio simulation package. For the calculations using WIEN2k code [5,6], the Perdew-Burke-Ernzerhof version of the generalized gradient approximation (GGA-PBE) was employed. For the calculations using VASP code, apart from the GGA-PBE methods, the hybrid exchange-correlation functional (HSE06) [7,8] was applied. For the calculations using CASTEP code [9], similar convergence parameters as used in VASP were employed.          Results were obtained using 128, 150 and 250 atoms system, using 2×2×2 k-points and Gamma point only. The black line is fitted from the experimental results by Ohta et al. [10]. The purple diamonds are the experimental results by Zhang et al. [11]. The red triangles are AIMD results from Pozzo and Alfe's work in 2016 [12]. The orange dash line is DFPT results taking saturation effects into consideration according to Cohen's data in 2018 [13]. The crosses are extrapolations to this density using the systematics of Stacey and Anderson based on the melting curve [14]. The green stars are experimental results by Bi et al. using shock-wave compression [15]. Guillot et al. [16].    x,y,z -y,x-y,z -x+y,-x,z -x,-y,z+1/2 y,-x+y,z+1/2 x-y,x,z+1/2 y,x,-z x-y,-y,-z -x,-x+y,-z -y,-x,-z+1/2 -x+y,y,-z+1/2