Machine Learning-Enabled Ab Initio Study of the Isotope Effect in SrTi^{18}O_{3}.

Area of Science:

• Condensed Matter Physics
• Materials Science
• Computational Materials Science

Background:

• Strontium titanate (SrTiO3) is a quantum paraelectric material exhibiting unique properties.
• The isotope effect, particularly oxygen substitution, is known to influence its phase transitions.
• Understanding these effects is crucial for designing advanced materials.

Purpose of the Study:

• To investigate the impact of oxygen-18 (18O) substitution on the quantum paraelectric properties of SrTiO3.
• To model the ferroelectric phase transition induced by 18O substitution using advanced computational methods.
• To determine the phase space boundaries between quantum paraelectric and ferroelectric states.

Main Methods:

• Utilizing the self-consistent harmonic approximation.
• Employing machine learning interatomic potentials for accurate atomic simulations.
• Performing temperature-dependent simulations incorporating quantum and anharmonic phonon effects.

Main Results:

• Calculations successfully reproduced the experimentally observed isotope effect, where 18O substitution induces ferroelectricity.
• A purely displacive description accurately modeled the ferroelectric phase transition in SrTiO3 with 18O.
• The phase space distinguishing quantum paraelectric SrTiO3 (16O) from ferroelectric SrTiO3 (18O) was found to be narrow.

Conclusions:

• Machine learning potentials enable accurate, temperature-dependent simulations of quantum and anharmonic effects.
• Oxygen isotope substitution plays a critical role in tuning the phase behavior of SrTiO3.
• The accuracy of electronic structure methods currently limits quantitative predictions of the SrTiO3 phase diagram.