A sharp look into tiny ferroelectric crystals

What happens to ferroelectric materials when their dimensions are greatly reduced? A team of researchers at HZB has now been able to show how this question can be answered in a detailed way.

Ferroelectric materials have a special inner structure. In the crystalline materials, ions align themselves differently within individual areas, the domains. This so-called polarisation can be changed or switched by electric fields or external pressure. These properties make ferroelectric materials interesting for various technical applications. For example, they are suitable as a material for capacitors - or, because the domains are very small, for storing large amounts of data in a small space.

But how do the ferroelectric properties change when the dimensions of the material are greatly reduced, for example to use them in nanoelectronic components? Experiments have shown that shrinking has enormous effects on the pattern of ferroelectric polarisation. “When the dimensions are reduced, the ferroelectric domains can take on a very different shape with a spatial extension of only several nanometers," explains Prof. Dr. Catherine Dubourdieu, head of the Institute Functional Oxides for Energy Efficient IT at the Helmholtz Zentrum Berlin für Materialien und Energie (HZB). "The diversity of electrical structures on a nanocrystalline scale opens up a whole new exciting horizon both for the understanding of the physics of these objects and for their potential applications. One key challenge is to be able to visualize such tiny domains in a non-destructive way.”

Catherine Dubourdieu and her team together with colleagues at Oak Ridge National Laboratory (ORNL) in the USA have now found a way to map the polarization pattern in thin ferroelectric layers precisely and non-destructively. To do this, the researchers relied on so-called contact Kelvin probe force microscopy (cKPFM) - a method that measures the material's electromechanical response under an electrical bias. To evaluate the big amount of data generated by mapping as low as 8x8 nm² pixel size, the HZB team applied a machine learning method. This made it possible to spatially resolve ferroelectric domains of less than 10 nanometres in size and of different polarization amplitudes. As sample material, the HZB researchers used a thin layer of barium titanate (BaTiO₃) in two crystalline forms: the so-called perovskite structure (one of the best-known ferroelectric materials) and the hexagonal structure, which is not ferroelectric at room temperature.

To check the reliability of the measurement method used, the HZB and ORNL teams also analysed the nanostructures using transmission electron microscopy (TEM). "The results of both experimental methods were in complete agreement," Dubourdieu is pleased to report. The scientists were also able to use this method to follow the ferroelectric pattern evolution while the sample was heated up to its paraelectric state. This opens up the possibility of also investigating the temperature dependence of the ferroelectric domain distribution and observing how ferroelectric domains form spatially below the so-called Curie temperature.

"Our results create a promising new perspective to study a large variety of polarization patterns at the nanoscale. This could lead, for example, to mapping the distribution of topological polar textures such as polar skyrmions which have been shown to have dimensions of about 10 nm. It could also be used to discriminate the polar domains from the non-polar ones in polycrystalline HfO₂-based ferroelectric thin films, a type of materials intensively studied for their potential integration in current nanoelectronics" says Dubourdieu. She adds “In the future, mapping ferroelectricity at the nanoscale with the help of machine learning will undoubtedly bring insights into phenomena occurring when dimensions are reduced and bring benefit for the integration of ferroelectrics into nanodevices.”

To the publication:

ACS Appl. Electron. Mater. (2021)

Sub-10 nm Probing of Ferroelectricity in Heterogeneous Materials by Machine Learning Enabled Contact Kelvin Probe Force Microscopy

Sebastian W. Schmitt, Rama K. Vasudevan, Maurice Seifert, Albina Y. Borisevich, Veeresh Deshpande, Sergei V. Kalinin, and Catherine Dubourdieu

doi: 10.1021/acsaelm.1c00569