Heterogeneous Transfer of Thin Film BaTiO$_3$ onto Silicon for Device Fabrication

TL;DR

The work advances integration of BaTiO3 electro-optic thin films with silicon by combining hybrid metal-organic MBE growth on a soluble SrO sacrificial layer with a polymer-free transfer via thermocompression bonding and chemical lift-off. The method delivers millimeter-scale, crack-free BaTiO3 on SiO2/Si while preserving crystallinity, enabling CMOS-compatible device fabrication and pattern transfer. The authors demonstrate lithography- and etch-compatible processing on the transferred film, highlighting its potential for scalable, silicon-based EO photonics. This heterogeneous transfer approach addresses known synthesis and integration challenges for BaTiO3, offering a practical path toward high-performance, silicon-based electro-optic devices.

Abstract

Thin film BaTiO$_3$ has one of the highest known Pockels coefficients (>1200 pm/V), making it an attractive material for use in electro-optic devices. It is advantageous to integrate BaTiO$_3$ on silicon to enable complementary metal-oxide-semiconductor (CMOS) compatible processing. However, synthesis of high-quality BaTiO$_3$ directly on silicon remains a challenge. Here, we synthesize BaTiO$_3$ using hybrid metal-organic molecular beam epitaxy (hMBE) and demonstrate its transfer onto silicon using thermocompression bonding and chemical lift-off. Hybrid metal-organic MBE enables self-regulated synthesis of highly stoichiometric thin films at high growth rates (>100nm/hr). Our transfer method results in millimeter-scale areas of atomically flat, crack-free BaTiO$_3$ making it a potentially scalable method. Finally, we demonstrate the applicability of our process to device fabrication through characterization of lithographically-patterned and etch-transferred sub-micron features.

Figures (10)

• Figure 1: Synthesis and characterization of thin film BaTiO$_3$. A. The tetragonal structure of a BaTiO$_3$ unit cell. The central atom, titanium, is slightly displaced in the $c$-axis direction, resulting in a ferroelectric polarization. B. We synthesize BaTiO$_3$ on (100)-orientated SrTiO$_3$ substrates with an intermediate sacrificial SrO layer (thickness not to scale). In situ reflection high energy electron diffraction (RHEED) patterns are shown for each step of the deposition, along the [100] substrate azimuth. C. AFM image of the BaTiO$_3$ thin film on SrO/SrTiO$_3$ displays atomic step terraces and a root mean squared roughness (RMS) of 153.3 picometers. D. The BaTiO$_3$ 002 peak in XRD indicates a $c$-axis oriented film. D. Full XRD pattern of BaTiO$_3$ grown on an SrO sacrificial layer and SrTiO$_3$ substrate.
• Figure 2: Transfer of BaTiO$_3$ onto silicon. A. We use hMBE to deposit strontium oxide (SrO) sacrificial layers and thin film BaTiO$_3$. B. The post synthesis stack with BaTiO$_3$ on SrO. C. SiO$_2$ is deposited onto the BaTiO$_3$ thin film and on a silicon substrate using CVD. D. Gold is deposited on top of the SiO$_2$ layers using electron beam evaporation. E. Thermocompression bonding with a force of 1000 N and 220° C is performed on the gold surfaces of BaTiO$_3$ and silicon. F. The stack is immersed in deionized (DI) water to dissolve the SrO sacrificial layer. G. The final stack has BaTiO$_3$ on a SiO$_2$ layer on a silicon wafer.
• Figure 3: BaTiO$_3$ transferred onto silicon. A. The XRD pattern of BaTiO$_3$ transferred onto a silicon substrate with the out-of-plane Bragg reflections (00$\textit{l}$) of BaTiO$_3$ showing that we have preserved the single crystalline structure after transfer onto silicon. B. XRD pattern of the 002 BaTiO$_3$ peak before transfer (maroon) and after transfer (pink). We note that there is a very slight shift in the peaks due to compressive strain from the bonding stack. C. The reciprocal space map (RSM) of BaTiO$_3$ 302 peak on silicon 103 peak shows a relaxed thin film in relation to the substrate. D. The AFM image of BaTiO$_3$ on silicon depicts atomic step terraces with a root mean squared roughness of 369 picometers. E. The scanning electron microscopy (SEM) image of BaTiO$_3$ on silicon shows a millimeter scale region that is flat and crack-free.
• Figure 4: BaTiO$_3$ fabrication tests. The SEM of nanopillars with feature sizes down to 1 $\mu$m is shown, demonstrating the BaTiO$_3$ is robust to standard nanofabrication.
• Figure S1: Peak Shifts after transfer of BaTiO$_3$ onto silicon. The zoomed in XRD of the BaTiO$_3$ 002 peak shows a big peak shift after transfer onto silicon. The lattice parameter increased from 4.057 $\textup{\AA}$ to 4.079 $\textup{\AA}$, indicating an in-plane compressive strain. Compared to our films that did not show such large peak shifts, we hypothesize that this shift is mainly due to the change in ramp rates for thermocompression bonding, which affected the thermal expansions of the different materials. In this film, we used a force ramp rate of 100 N/s and a temperature ramp rate of 20° C/s while in the films that did not have drastic peak shifts we lowered our ramp rates to a force ramp rate of 6 N/s and a temperature ramp rate of 1° C/s. Using slower ramp rates should not only help with reducing peak shifts but in thicker films this will also avoid instances of the thin film cracking as we cycle through its Curie temperature of 130° C. One component we have not accounted for is that this was a much thinner film (30 nm) compared to later films that did not have larger peak shifts (100 nm). So it could be possible that the thickness also played an important role in peak shifts.