Laser Wavelength Dependence on Perovskite Interface Elemental Diffusion During Atom Probe Experiments

Among perovskite metal oxide of chemical formular ABO3, material interfaces with different B-site cations, such as SrTiO3 (STO) and LaMnO3 (LMO) are highly desired for spintronic applications [1]. Previous studies on STO/LMO superstructures showed enhanced magnetism per Mn ion compared to bulk LMO films for specific periodicities [2]. Scanning transmission electron microscopy (STEM) annular bright field imaging (ABF) has shown perturbations in the LMO oxide octahedral tilt angle within 3-4 unit cells at the LMO/STO interface [3]. This is consistent with the LMO and STO layer sizes that show the most enhanced LMO magnetism in STO/LMO superstructures. SrTiO3(STO)/LaMnO3(LMO)/La(Al,Sc)O3(LASO) superstructures with different growth orders (STO→LMO→LASO→STO… and STO→LASO→LMO→STO…) were grown by atomic layer deposition (ALD) to assess the effect of the LMO/STO termination plane on the magnetic properties (see Figure 1 for the sample schematic). The LASO layer was used to isolate the single STO/LMO interfaces because the LASO/LMO interface has no effect on the material’s magnetism. The STO→LMO (LMO on top of STO (LMO-OTO-STO)) sample consistently showed higher magnetism than the LMO→STO (STO-OTO-LMO) samples. STEM-ABF imaging and EELS did not show any noticeable differences between the samples. STEM-electron energy loss spectroscopy (EELS) data can be difficult to interpret at the atomic scale because of channeling effects and is only sensitive to approximately 1 at. %, and therefore, atom probe tomography (APT) was used to assess chemical diffusion across the interfaces due to its higher chemical sensitivity and close to Angstrom z resolution. Most of the APT experiments were performed using a LEAP4000XHR equipped with a 3.5 eV or ∼355 nm laser. Sample yield was less than 10%, and only small volumes were collected. The results showed Mn diffusion into STO and Ti diffusion into LMO for the STO-OTO-LMO sample (sample with worse magnetic properties), while Mn nor Ti diffusion was not found in the LMO-OTO-STO sample. The STO-OTO-LMO sample was also run using a LEAP6000XR equipped with a 4.8 eV or ∼257 nm laser. Three full superstructures were captured out of three attempts (100% yield) and much larger volumes were collected. Surprisingly, the results only showed Mn diffusion into STO for the STO-OTO-LMO sample, which was inconsistent with the LEAP4000XHR results. Repeat LEAP4000XHR experiments using backside sample preparation indicated Ti retention during the APT experiment creating a false Ti diffusion profile into LMO. Ti retention for the LEAP4000XHR experiments can be explained by STO’s band gap and the larger laser powers needed for field evaporation with a 3.5 eV vs. 4.8 eV laser (100 pJ vs. 10 pJ, respectively). STO has a 3.25 eV band gap at 4K and strong optical absorption coefficient for 3.5 eV and 4.8 eV photons, and therefore, both laser energies heat the sample effectively [4]. However, the 4.8 eV laser can excite STO’s direct band gap, while the 3.5 eV laser only has access to STO’s indirect band gap [5]. The purer field evaporation behavior of the 4.7 eV excitation reveals that laser assisted insulator field evaporation is strongly dependent on direct band gap excitation, while heating through optical absorption is secondary. The comparison in APT data quality and mechanisms for differences in the field evaporation behavior for STO/LMO/LASO superstructures will be discussed. Overall, the APT results show that the lower STO-OTO-LMO magnetism is consistent with Mn diffusion into the STO layer during growth, which was unable to be revealed definitively using STEM techniques [6]. (a) The sample superstructure schematic showing the growth order. The interesting interface for magnetic properties is highlighted. (b) Direct comparison of data collected from the LEAP4000XHR (top) and LEAP6000XR (bottom), which are displayed at the same magnification (the scale bar is relevant for all datasets). The samples always fractured before reaching the STO substrate using a 3.5 eV laser, while all runs were stopped manually when in the STO substrate using a 4.8 eV laser.