Self-assembled vertically aligned nanocomposite systems integrated on silicon substrate: Progress and future perspectives

Functional oxide thin films have aroused enormous research interest in recent decades owing to their full spectrum of electronic, magnetic, and optical properties such as semiconducting, dielectric, superconducting, (anti)ferroelectric, (anti)ferromagnetic, multiferroic, nonlinear optical effects, etc., which open up tremendous opportunities in nanoelectronics and energy devices such as nonvolatile random-access memory (RAM), superconducting device applications, and energy conversion and storage devices.^1–7 Apart from single-phase oxide films, nanocomposite thin films that combine two or more material components into one framework exhibit much more versatile and intriguing physical properties that cannot be achieved by either phase. For example, ferroelectric BaTiO₃ (BTO) and ferromagnetic CoFe₂O₄ (CFO) have been coupled to form a self-assembled multiferroic BTO–CFO nanostructured thin film;⁸ a wide range of secondary phases have been introduced in La_0.7Sr_0.3MnO₃ as VAN structures for enhanced low-field magnetoelectric properties that could not be easily achieved in single-phase LSMO thin films;^9–12 a novel vertical ionic conducting channel has been introduced in cathode-electrolyte VANs for significantly enhanced solid oxide fuel cell power performance.^13–15 

Different from the typical “top-down” approaches for nanocomposite film fabrication such as electron-beam lithography (EBL) and focused ion beam (FIB), which involve multiple steps such as templating, coating, patterning and etching, etc.,^16–19 the pulsed laser deposition (PLD) has been demonstrated as a simplified “bottom-up” fabrication technique for growing complex oxide-based nanocomposite films through a one-step growth method.^20–23 The targets for VAN growth are typically prepared via a solid-state sintering process by fully mixing two material components with a predefined molar ratio and then ablated by a high-energy laser beam to evaporate two kinds of materials for film deposition. Owing to the different surface energy and wettability between two phases and a substrate, one oxide phase presents the layer-by-layer growth mode (Frank–Van der Merwe) to form the film matrix, while the other phase performs the island growth mode (Volmer–Weber) and becomes secondary nanopillar or nanocolumn phase embedded in the matrix, forming the self-assembled VAN morphology.^24–27 The VAN thin films exhibit a strongly enhanced interfacial strain coupling effect and highly tunable multifunctionalities such as ferroelectric, ferromagnetic, multiferroic, magnetoresistance (MR), magnetoelectric (ME), magneto-optical (MO), superconductivity, etc.^{8,22,25,28,29} Various two-phase oxide-oxide film systems have been explored and studied including BiFeO₃–Sm₂O₃ (BFO–SmO),^22,30 La_0.7Sr_0.3MnO₃ (LSMO)–ZnO,⁹ BTO–SmO,²³ PbTiO₃–CFO,³¹ LSMO–MgO,³² BaZrO₃–YBa₂Cu₃O_7−x,³³ etc. Recently, oxide-metal nanocomposite films have emerged as a new category of functional thin films that exhibit an extraordinary optical, magnetic, and magneto-optical coupling effect by incorporating plasmonic metals (e.g., Au and Ag) or ferromagnetic metals (e.g., Fe, Co, and Ni) into the oxide matrix.^34–41 

Nevertheless, most of the previous thin film studies focus on the films grown on single-crystal oxide substrates,^42,43 i.e., SrTiO₃ (STO), LaAlO₃ (LAO), MgO, and sapphire $(α−Al2O3)$⁠, which are costly and relatively small in dimension, therefore, not suitable for large-scale integration toward device applications. Silicon (Si), as the backbone for modern semiconducting industry, is an ideal substrate for future integration of oxide-based nanoelectronics devices with complementary metal-oxide-semiconductor (CMOS) devices.^44,45 Hence, it is highly desirable to grow the multifunctional VAN thin films on Si substrates by maintaining their promising properties and performances. However, the direct growth of functional oxides on Si fails to achieve highly epitaxial film growth due to the oxidation of Si surface during deposition at high temperatures and the large lattice mismatch between the oxide film and Si substrate. One effective solution is to grow a set of buffer layers on Si to provide structural compatibility and thermal and chemical stabilities between the deposited films and underlying substrates.^46–50 Different buffer layers have been demonstrated including pure STO,^46,47 CaTiO₃,⁴⁸ TiN/STO,^51,52 Sr(Ti_1−xFe_x)O₃/CeO₂/YSZ,⁴⁹ MgIn₂O₄/CeO₂/YSZ,⁵³ and (La,Sr)CoO₃/CeO₂/YSZ,⁵⁴ etc. Among all these buffer layer candidates, bilayer STO/TiN buffer stack has been proven to be successful in enabling multiple VAN systems growth on the Si substrate because of their chemical and thermal stabilities as well as the high epitaxy quality that facilitates the epitaxial growth of the nanocomposite film.^{51,52,55–57}

As the 3D schematic illustrations depicted in Fig. 1(a), STO/TiN buffer layer stack is applied for both oxide-oxide and oxide-metal VAN thin films growth on Si. The lattice matching model shown on the right shows the domain matching epitaxy (DME) mode of TiN on Si (4TiN: 3Si).^51,55–57 STO is implemented as the second buffer layer to further facilitate the epitaxial growth on oxide-based nanocomposite film as it has a closer matched lattice parameter and better wettability with most oxides. In contrast, for the deposition of nitride-metal VAN film shown in Fig. 1(b), as the nitride phase (i.e., TiN) already serves as the film matrix, no further buffer layer is needed during the film growth and thus a direct VAN growth on Si is possible. In this review, all the VAN systems integrated on the Si substrate are summarized in Sec. II. Several oxide-oxide (Sec. III), oxide-metal (Sec. IV), and nitride-metal (Sec. V) VAN films grown on Si are reviewed, exhibiting highly anisotropic and tunable physical properties. In Section VI, future perspectives and remaining challenges of VAN integrated on Si towards magnetic and photonic device applications are discussed. Tunable properties of the VAN system as well as the robustness and scalability of Si substrate show great promise in the next generation of nanoelectronics and nanophotonic device applications.

Table I summarizes the crystal structures and lattice parameters of common oxides, nitrides, and metal components that have been explored in VAN systems. It can be seen that the lattice constants of most oxides are between 3.7 and 4.2 Å, while some other oxides with larger unit cells (i.e., CeO₂, Sm₂O₃, etc.) conduct the 45° in-plane rotation then follow the cube-on-cube or DME during film growth.^10,23,30,62 On the other hand, nitrides usually have larger lattice parameters ranging from 4.2 to 4.5 Å, showing a good DME with Si (a = 5.431 Å) as discussed before. The most common metal candidate used for the oxide-metal VAN system is Au (a = 4.065 Å), owing to its excellent chemical stability and the closer lattice constant with oxides resulting in smaller lattice mismatch and less strain and defect formation during the film growth. Other metals such as Fe and Co have also been incorporated into some VAN systems showing distinct ferromagnetic responses, which will be discussed in great detail later.

All the VAN systems that have been demonstrated on Si are summarized in Table II. It can be seen that perovskite oxides have served as the basic platform for VAN integration coupled with other crystal structured components such as rocksalts, fluorites, and spinels for the study of strain control and multifunctionality coupling. Different from the single-phase oxide films where the lattice strain only results from the lattice mismatch between the epitaxial film and substrate, the VAN systems combine both in-plane strain between the film and substrate as well as out-of-plane strain between two immiscible phases, leading to much more complex and tunable strain states within VAN systems. Hence, the lattice parameter and crystal structure are the top factors to be considered during materials selection because they directly impact the strain and defects as well as the epitaxial quality of the as-deposited thin films.

Oxide-oxide nanocomposite thin films are the first explored VAN systems due to the versatility of functional oxides family and their excellent chemical and thermal stability. Here, three oxide-oxide VAN systems integrated on Si, i.e., LSMO–ZnO,⁵¹ BTO–CeO₂,⁵² and LSMO–NiO,⁷⁶ are reviewed, showing tailorable magnetoresistance (MR), ferroelectricity, and exchange bias (EB) effect, respectively. The integration of multifunctional oxide-oxide VAN films on Si is a critical step for the application of VAN in future spintronic devices.

La_0.7Sr_0.3MnO₃ (LSMO), a well-known ferromagnetic perovskite oxide showing fascinating colossal magnetoresistance (CMR), low-field magnetoresistance (LFMR), magnetic anisotropy, and spin-glass like behaviors, etc., has become a promising candidate in the application of spintronic devices.^79–81 Because the MR is strongly controlled by grain boundaries (GBs) and interfacial coupling effect,⁸² the LSMO–ZnO nanocomposite films demonstrated on both STO and Si substrates exhibit enhanced LFMR effect and anisotropic magnetic properties.^9,51 As shown in Fig. 2(a), the low-magnification bright-field (BF) TEM image shows the growth morphology of self-assembled LSMO–ZnO nanocomposite film grown on the STO/TiN buffered Si substrate.⁵¹ The selected area electron diffraction (SAED) pattern shown as the inset in Fig. 2(a) indicates the clear phase separation of LSMO and ZnO within the film. High-resolution TEM image [Fig. 2(b)] of the interface area between the buffer layer and substrate shows an ultrathin SiO₂ layer (∼3–5 nm) formed on the Si surface due to the minor oxidation of Si at high temperature before deposition. The high-resolution TEM (HRTEM) image shown in Fig. 2(c) reveals the phase separation and high-quality epitaxial growth of both LSMO and ZnO nanocolumns, which are labeled out based on their out-of-plane lattice parameters. Figure 2(d) illustrates the plan-view TEM image of the LSMO–ZnO film, showing the ordered alternate growth of LSMO and ZnO domains.

A more detailed study was conducted to investigate the LFMR of LSMO–ZnO (L–Z) films with different ZnO composition ratios. As shown in Fig. 2(e), the resistivity ratio $(ρH/ρ0)$ of a series of L_1−xZ_x composite films (x = 0, 0.3, 0.5, 0.7) was measured at 80 K with an external magnetic field ranging from −1 to 1 T. It can be seen that all resistivity ratio curves for all L_1−xZ_x films exhibit a sharp drop at low magnetic fields (0–0.5 T), then followed by a more gradual drop at higher magnetic fields (0.5–1 T). It is clear that the film with the 70% ZnO molar ratio has the maximum resistivity, which is due to the increase of artificial grain and phase boundaries that suppress the ferromagnetic (FM) double-exchange interaction between neighboring FM nanodomains.^83,84 The MR ratios as a function of temperature of all the film samples, defined as $MR=(ρ0−ρH)/ρ0$⁠, are plotted in Fig. 2(f). It is obvious that the MR (%) of all the samples increases as the measuring temperature decreases from room temperature (RT) to low-temperature range, where L_0.3Z_0.7 (70% ZnO molar ratio) film shows the maximum MR ∼ 55% at low temperature. It is also noted that the peak MR value (∼32%) of the L_0.5Z_0.5 film grown on the buffered Si substrate is very close to that of the L_0.5Z_0.5 film deposited on STO (001) (∼30%) in the previous study,⁹ which also indicates the successful integration of the nanocomposite film on Si through buffered growth. Overall, the incorporation of nonmagnetic insulating ZnO phase contributes to creating more artificial grain and phase boundaries, acting as insulating tunnel barrier and inducing local spin disorder,⁸⁵ leading to the enhanced and tunable LFMR effect.

BaTiO₃ (BTO) is a perovskite material, which has consistently drawn significant attention because of its dielectric, ferroelectric, piezoelectric, and nonlinear optic properties and potential applications in dynamic random-access memories (DRAMs), nonvolatile memories (NVRAMs), integrated devices, and optical modulators.^86–89 The main drawback for BTO as ferroelectrics is that it suffers from high dielectric losses, which can be effectively reduced by doping or addition of a secondary low dielectric constant material. Therefore, CeO₂, as a rare-earth fluorite oxide, has been considered as a promising secondary phase in BTO.⁹⁰ Here, a self-assembled BTO–CeO₂ VAN film is integrated on STO/TiN buffered Si.⁵² Owing to different lattice parameters of BTO (a = 3.994 Å) and CeO₂ (a = 5.431 Å), BTO presents the cube-on-cube, while CeO₂ shows a 45^o in-plane (IP) rotation growth. Figure 3(a) shows the low-mag cross-sectional TEM image of the BTO–CeO₂ film grown on STO/TiN buffered Si. The corresponding SAED pattern is shown as an inset, indicating the matching relationship to be $(011)BTO//(010)CeO2//(011)STO,$ confirming the IP rotation of CeO₂. The high-angle annular dark field mode (HAADF) STEM image of the film is shown in Fig. 3(b). Since the intensity of the HAADF-STEM image is proportional to Z^1.7,^91–93 the CeO₂ phase exhibits a brighter contrast than that of BTO, as labeled in Fig. 3(b). The HRTEM image shown in Fig. 3(c) shows the good epitaxial growth of TiN and STO buffer layers on Si where the sharp STO/TiN and TiN/Si interfaces are identified indicating no obvious interdiffusion between the phases. Figure 3(d) shows the 7 nm SrRuO₃ (SRO) epitaxial layer, which is deposited on top of the STO buffer to be the bottom electrode for the ferroelectricity measurement. The high-quality epitaxial growth of both BTO and CeO₂ phases is confirmed by the clear phase boundaries (labeled with curved dashed lines) and the average nanocolumns width is around 3–5 nm.

To investigate the impact of buffer layers on the nanocomposite film growth quality, the P–E measurements for the BTO–CeO₂ films deposited on Si with different buffer layers are conducted, as plotted in Fig. 3(e). It is clear that the sample without buffer layers (black line) shows no ferroelectric response and the hysteresis loop is in characteristic of the lossy dielectric. The film grown on the pure STO buffer on Si (yellow-green line) also demonstrates a ferroelectric response but not as good as the sample grown on STO/TiN buffered Si. It can be explained by the less ordering in the STO buffer layer due to a large lattice mismatch with Si so that may result in some local textured growth and degrade the VAN film quality. Besides buffer layers, the deposition frequency also plays an important role in the epitaxial growth of the thin film. Here, a series of BTO–CeO₂ film samples grown on the STO/TiN buffered Si substrate are deposited at 2, 5, and 10 Hz, respectively. Based on their θ–2θ XRD spectra, the c-lattice values of the 2, 5, and 10 Hz film samples are calculated to be 4.06, 4.06, and 4.07 Å, and the out-of-plane tensile strains are calculated to be 0.52%, 0.68%, and 0.81%, respectively. Moreover, the full width at half maximum (FWHM) for the BTO (001) peak is measured to be 0.329°, 0.293°, and 0.36°1 for the 2, 5, and 10 Hz samples, respectively.⁵² Therefore, it is expected that the 5 Hz grown sample has more consistent c-lattice and better ferroelectric response since it is strongly correlated with the c-lattice polarization under the electric field. As shown in Fig. 3(f), the 5 Hz deposited sample shows the most distinct ferroelectric P–E loop, which is consistent with expectation.

The EB effect is the phenomenon that the magnetization hysteresis curve shifts from the origin due to the spin interactions between FM and antiferromagnetic (AFM) layers.^94,95 EB has been widely used in many application fields such as magnetic memory, magnetic field sensors, and recording heads for hard disk drives (HDD).^96,97 In this study, FM LSMO and AFM NiO are selected as a novel FM-AFM VAN system for the demonstration of EB oxides integrated on the Si substrate.⁷⁶ The SRO/TiN buffer layer stack is adapted to facilitate the epitaxial growth of the LSMO–NiO VAN film. Figure 4(a) shows a low-mag cross-sectional TEM image of the (LSMO)_0.25(NiO)_0.75 film. Two buffer layers, i.e., TiN and SRO are clearly identified and the nanocomposite film exhibits the VAN morphology. Based on the inserted SAED pattern, the epitaxial orientation relationship is determined to be LSMO (002)//NiO (002)//SRO (002) and LSMO [200]//NiO [200]//SRO [200]. The HRTEM image at the LSMO/NiO interface shown in Fig. 4(b) demonstrates the high epitaxial growth and consistent epitaxial matching relation between LSMO and NiO nanocolumns. The HAADF-STEM image and corresponding EDS mapping results shown in Figs. 4(c) and 4(d) further demonstrate the well separated LSMO and NiO phases without obvious interdiffusion. It can be seen that LSMO and NiO follow the VAN growth morphology to form nanocolumns with the average width of ∼ 10–20 nm.

To explore the EB effect of the LSMO–NiO (L_1−x− N_x)VAN films, magnetization hysteresis measurement has been carried out for all the film samples with different NiO molar ratios (%). The pure LSMO film was also measured for comparison. Both out-of-plane (OP) and IP hysteresis loops show the horizontal shifts in the L_1−x− N_x samples. By defining $HEB=|H++H−|/2$ (where $H+$ and $H−$ represent the positive and negative values of coercivity as the magnetization goes to zero), the H_EB values of all the L_1−x− N_x samples (x = 0, 0.25, 0.5, 0.75) are plotted in Fig. 4(e). It can be seen that both OP and IP H_EB values follow the same trend that the H_EB value increases by increasing the NiO composition ratio because of the more vertical FM/AFM interface coupling effect so that the L_0.25N_0.75 sample exhibits the highest H_EB value while no shift is found for the pure LSMO film. It is also noted that the OP H_EB components for different samples all show larger values than the corresponding IP H_EB, and the difference becomes larger with increasing the NiO ratio. This is mainly because the magnetic spins within the AFM NiO phase are aligned along the $[112¯]$ direction under the magnetic field,⁹⁸ where the OP spin component is larger than the IP spin component, resulting in the larger OP H_EB values for all the samples. Furthermore, coercivity (H_C) values of all the samples are calculated by $HC=|H+−H−|/2$ and plotted in Fig. 4(f). The general trend shows that the OP H_C is higher than IP H_C, which further confirms the anisotropic nature of the L_1−x− N_x samples due to the larger spin alignment along the OP direction. Therefore, the LSMO–NiO nanocomposite film integrated on Si exhibits obvious EB effect and enhanced coercivity owing to the strong coupling effect between FM LSMO and AFM NiO along the vertical interfaces. This study has paved the avenue of Si-integration of FM-AFM nanocomposite films for enhanced EB effect and coercivity toward magnetic storage device applications.

Besides the well explored oxide-oxide VANs, oxide-metal VAN thin films have emerged as a new class of nanocomposite thin films in recent years showing distinct ferromagnetic and plasmonic properties owing to the Fe, Co, Ni, Au, and Cu inclusions. Vertically aligned magnetic nanostructures embedded in nonferromagnetic matrices, for example, could offer an attractive framework for exploring anisotropic materials and properties with potential applications in devices such as magnetic tunnel junctions (MTJ).⁹⁹ A variety of oxide-metal VAN systems have been explored until now, such as BTO–Au,^34,37,40,100 BTO–Fe,³⁸ BZO–Co,^36,99 ZnO–Au,¹⁰¹ BTO–Au_xAg_1−x,¹⁰² BTO–ZnO–Au,^103–105 LSFO–Au–Fe,⁴¹ etc., while very limited oxide-metal film systems have been successfully integrated on Si substrate due to the large lattice mismatch and oxidation of Si at high temperatures. Here, LSFO–Fe,⁵⁵ BTO–Fe,⁵⁶ and BTO–Au,⁵⁷ are the only oxide-metal VAN film systems that have been demonstrated on Si up-to-date and thus are reviewed. More efforts are obviously needed for future oxide-metal VAN integration on Si toward scalable electronic and photonic device applications.

Practically speaking, the growth of oxide thin films is typically performed in an oxygenated environment at temperatures greater than 400 °C—conditions under which most metals tend to oxidize, therefore negating any effort to create a purely metal phase. Nonetheless, in 2004 Mohaddes-Ardabili et al. demonstrated a self-assembled oxide-metal VAN by depositing La_0.5Sr_0.5FeO₃ in a reducing environment (i.e., high vacuum), which resulted in its decomposition into Fe metal nanopillars embedded within an LaSrFeO₄ (LSFO) oxide matrix.¹⁰⁶ This decomposition-based oxide-metal VAN showed major changes both in its nanostructure and physical properties, featuring prominent structural and magnetic anisotropy. Inspired by this work, the LSFO–Fe system is chosen as a pioneering oxide-metal system to be deposited on STO (001) substrate under high vacuum.¹⁰⁷ Next, using the STO/TiN bilayer buffer technique which was shown to be effective for the integration of various oxide-oxide VANs on Si, LSFO–Fe VAN has been integrated on Si substrates as a first step toward some of the proposed device applications.⁵⁵ A schematic illustration of the Si substrate, buffer layer (STO/TiN), and LSFO–Fe VAN stack is shown in [Fig. 5(a)]. A cross-sectional EDS elemental map of the stack in [Fig. 5(b)] emphasizes that the TiN and STO buffer layers are distinctly present, and the LSFO–Fe nanostructure is like that of a VAN. In fact, the morphology of the LSFO–Fe film integrated on Si is similar to those same composition films grown on STO substrates in the same study as well as in previous studies.^55,106,107 The oxide-metal VAN morphology is also apparent in the magnetic properties of the films. In Figs. 5(c) and 5(d), hysteresis loops of LSFO–Fe films grown on STO and STO/TiN/Si substrates both exhibit a similar ferromagnetic behavior in terms of saturation magnetization, coercivity, and a preferred magnetization along the out-of-plane orientation, the latter being attributed to the perpendicularly anisotropic Fe nanostructures in the film.

Overall, the magnetic anisotropy achieved by the LSFO_0.7:Fe_0.3 thin film on Si (001) is significant and opens a new avenue for easy integration of magnetic oxide-metal systems on Si for ferromagnetic tunnel junctions, high-density magnetic data storage, and other spintronic devices. However, it is important to note at least one major drawback of oxide-metal VANs formed by decomposition (e.g., LSFO–Fe), and that is there are only a few known materials that decompose into separate oxide and metal components. This material restriction, which severely limits the exploration of other combinations of oxides and metals to form VAN, together with the exacting deposition conditions necessary for the decomposition, may help explain why there have been so few studies on oxide-metal VAN systems achieved by decomposition, particularly the LSFO–Fe system.

Different from the decomposition-based VAN growth, the recent cogrown oxide-metal VANs demonstrate great potentials in versatile materials selections and morphology designs. Specifically, a high vacuum or reducing deposition environment is adopted during growth to limit the oxidation of non-noble metals, even those that are particularly susceptible to oxidation such as Co or Fe. The BTO–Fe is one of these oxide-metal systems which is of particular interest due to the combination of BTO, a well-studied ferroelectric, together with Fe, one of the strongest ferromagnetic elements, to form a room-temperature multiferroic material. Here, BTO–Fe nanocomposite thin film was integrated on Si shortly after the original report to demonstrate its reproducible multiferroic property in Si-based devices.⁵⁶ The schematic configuration of the BTO–Fe film with STO/TiN stack buffer layers is shown in Fig. 6(a). The piezoresponse force microscopy (PFM) phase and amplitude switching curves plotted in Fig. 6(b) confirm the ferroelectric property of the nanocomposite film. The EDS elemental mappings of Fe, Ba, and Ti shown in Figs. 6(c)–6(e) provide direct evidence of the bilayer buffer and the BTO–Fe film where Fe nanopillars are embedded in the BTO matrix. Interestingly, despite being grown on three different buffer configurations—a STO/TiN bilayer buffer, TiN only buffer, and no buffer—all samples exhibit ferromagnetic behavior and perpendicular anisotropy, as indicated by the IP and OP hysteresis loops in Figs. 6(f) and 6(g). Nanostructure analyses suggest that this is due to their shared morphology consisting of Fe nanopillars in a BTO matrix, similar to that which was observed in LSFO–Fe VANs. Based on the ferroelectricity and ferromagnetism measurement results, the room-temperature multiferroic property of the BTO–Fe VAN film is confirmed. Overall, the demonstration of BTO–Fe thin films on Si using buffer layer configurations presents an effective approach for integrating room-temperature multiferroic materials on silicon.

Oxide-noble metal VANs (i.e., BTO–Au, BTO–AuAg, Zno–Au, etc.) have been integrated on oxide substrate showing typical plasmonic features in the visible to near-infrared wavelength region.^{34,37,39,101,102} The use of noble metals provides a unique advantage for the oxide-metal VAN growth in that they are highly resistant to oxidation, even in the deposition conditions necessary for the oxide thin film growth, therefore allowing for many more combinations of oxides and metals to be explored through the VAN framework. As a step toward future Si-based photonic devices, BTO–Au was integrated on Si substrates a few years following the initial report of the material system.⁵⁷ Integration of the BTO–Au VAN was achieved using the same STO/TiN buffer stack grown in ultrahigh vacuum (UHV), as schematically shown in Fig. 7(a). The EDS mapping result in Fig. 7(b) demonstrates the typical VAN morphology of the film. In comparison, the same target was used to grow the BTO–Au film on STO/TiN buffered Si in 40 mTorr O₂ atmosphere. The TEM and EDS mapping results show that the film maintains the VAN morphology, as shown in Fig. 7(d). Further analysis of the interface reveals that a thin amorphous layer of SiO₂ formed at the silicon and buffer layer interface due to the minor oxidation of TiN and Si under the oxygen partial pressure at high temperature, although the growth quality of the BTO–Au was not severely impacted. More interestingly, by comparing the microstructure analysis results of the BTO–Au grown in UHV and 40 mTorr O₂, it is found that the size and distribution of the Au nanopillars in the BTO matrix changed significantly, suggesting that Au nanopillar geometries can be tuned by oxygen partial pressure control during the film growth.

Owing to the metallic Au nanopillars embedded within the dielectric BTO matrix, BTO–Au VANs are capable of exhibiting highly anisotropic optical properties. Thus, the IP and OP permittivity values for the films were derived separately by fitting the spectroscopic ellipsometry data. Figures 6(e) and 6(f) show the real and imaginary permittivity values of BTO–Au films grown on Si deposited in UHV and 40 mTorr O₂, where both curves exhibit the Au plasmonic absorption peak position at around 600 nm as the arrows marked out. Due to the Au nanopillars being oriented perpendicular to the surface of the film, the real permittivity OP components $(ε⊥′)$ of both BTO–Au films decrease from positive to negative, showing the epsilon-near-zero (ENZ) feature in the Vis-NIR wavelength region. On the other hand, the real permittivity IP components $(ε∥′)$ of both films remain positive throughout the whole wavelength range mainly due to the dielectric BTO matrix. The opposite signs of OP and IP real permittivity components satisfy the conditions for Type I hyperbolic dispersion $(ε∥′>0,ε⊥′<0)$⁠.¹⁰⁸ Representative k-space isofrequency surfaces of the BTO–Au films illustrate how the change in real permittivity values transform the isofrequency surfaces from an ellipsoid to a hyperbola, as the 3D insets plotted in Fig. 7(e). More importantly, it is noted that the ENZ wavelengths conduct a blueshift from 1096 nm for the BTO–Au film grown in UHV to 786 nm for the film grown in 40 mTorr O₂, indicating that the BTO–Au film grown in 40 mTorr O₂ exhibits a more metallic behavior due to the higher Au nanopillar density as revealed by the TEM and EDS results [Figs. 7(a)–7(d)]. Therefore, the different growth conditions result in variable diameter and density of the Au nanopillars within the hybrid thin films, leading to highly tunable optical responses such as hyperbolic dispersion in the Vis-NIR regime.

The integration of various oxide-metal VAN thin films on silicon highlights many important aspects. For one, the critical role of buffer layers for epitaxial growth of VANs on Si substrates is extended beyond oxide-oxide systems to the oxide-metal ones. Second, of the three main methods to fabricate oxide-metal VANs, all of them were shown to be compatible with the buffer layers and integration on Si. Lastly, the compatibility of these unique materials and properties with a popular semiconducting material such as Si shows great promise toward oxide-metal VAN implementation in devices.

Beyond the abovementioned oxide-oxide, oxide-metal VAN systems, in recent years nitride-metal VAN thin films have also been successfully integrated on various single-crystal substrates such as STO and MgO.^109–112 The typical transitional metal nitrides, i.e., TiN, TaN, and HfN, show strong surface plasmon and robust thermal stability, which enable them as good matrix material candidates for the nitride-metal VAN system integrated on Si substrate towards applications of future nanophotonic devices. In a recent study, a 3D plasmonic framework of multilayered (ML) Au–TaN/Au–TiN × N (N = 1, 2, 3, 4) stack thin film is designed and fabricated.⁷⁸ Figures 8(a)–8(d) show the HAADF-STEM images of the series of Au–TaN/Au–TiN, Au–TaN/Au–TiN × 2, Au–TaN/Au–TiN × 3, and Au–TaN/Au–TiN × 4 multilayered VAN thin films. Owing to the DME growth of TiN on Si substrate as discussed before, all the films show good epitaxy quality grown on Si (001) substrates without buffer layers. Since Ta has a higher Z number than that of Ti and the HAADF-STEM image intensity is proportional to atomic number (Z),^91–93 the Au–TaN phases show a brighter contrast as seen in Fig. 8(a)–8(d).

As for the optical properties of this 3D ML structure, owing to the vertically grown Au nanopillars, all the Au-nitride multilayer films exhibit hyperbolic dispersion in the Vis-NIR regime. Figure 8(e) shows the real permittivity (ε′) curves of all the ML films in the wavelength range from 300 to 1500 nm. It is noted that the IP permittivity components $(ε∥′)$ are all positive within the whole wavelength range, while the OP permittivity components $(ε⊥′)$ decrease from small positive to large negative values. It can be explained by the existing vertical Au nanopillars within the matrix that lead to the negative permittivity along the OP direction, leading to the hyperbolic behavior $(ε∥′⋅ε⊥′<0)$ of the metamaterials.^29,37,39,57 Moreover, the enlarged permittivity ENZ region of all the ML samples is shown in Fig. 8(f). It is obviously seen that the ENZ wavelengths conduct a blueshift decreasing from ∼605 nm for Au–TaN/Au–TiN bilayer film to ∼405 nm for Au–TaN/Au–TiN × 4 Ml film, indicating the gradual free electron concentration increase by increasing the number of stacking layers within the Au–TaN/Au–TiN ML structure. Therefore, the hybrid ML VAN structure presents a novel way of achieving optical property tuning for hybrid metamaterials.

Very recently, a novel nitride-oxide system, where plasmonic TiN coupled with AFM NiO to form the VAN structured film, has been integrated on the single-crystal MgO substrate, showing hyperbolic dispersion and magneto-optical Kerr effect (MOKE).¹¹³ A following work has been done on a nitride-oxide-metal system—TiN–NiO–Au, exhibiting core-shell nanopillar structure and ordered distribution as well as tunable FM and MO responses.¹¹⁴ Hence, more nitride-based VAN systems still need to be explored and integrated on Si for promising tunable and modulated optical-based devices.

In this review, the current progress of self-assembled oxide-oxide, oxide-metal, and nitride-metal VAN systems integrated on the Si substrate has been reviewed and discussed. Buffer layered growth is an effective approach for obtaining epitaxial film growth on Si which has been applied in various oxide-oxide and oxide-metal VAN systems, while no buffer layer is needed for nitride-metal VAN film growth due to the close lattice parameters between typical nitrides and Si. Owing to the versatile materials system selection, the Si-integrated VAN films exhibit a strong interfacial strain coupling effect as well as enhanced properties such as MR, ferroelectricity (FE), ferromagnetism (FM), EB, and optical properties. Through tuning the materials components ratio or growth parameters such as deposition frequency and background oxygen pressure, the growth morphology and corresponding physical properties of the VAN films can be effectively tailored. Future endeavors are needed for materials system exploration, physical properties tuning and growth optimization toward future Si-integrated nanoelectronics and nanophotonic device applications. To be more specific, some of the potential device applications and the needs of new VAN systems integrated on Si are discussed in detail below.

MTJs, consisting of two layers of magnetic metal (i.e., CoFe, CoFeB, and NiFe, etc.) separated by an ultrathin layer of insulator, have shown promising applications in data storage devices such as magnetic random-access memory (MRAM) which has a great potential for next-generation high-density nonvolatile memory and logic chips.^115–117 Therefore, tremendous work has been devoted to developing new oxide-based MTJs exhibiting good thermal stability, low-current induction, and high MR effect.^118–120 

In a recent work of self-assembled Co–BaZrO₃ VAN design for memory device integration, a new type of MTJ structure is proposed, as the concept device structure shown in Fig. 9(a).⁹⁹ As illustrated in Fig. 9(b), a bottom FM LSMO layer (blue phase) was first deposited, followed by an ultrathin BZO insulating layer (the pink thin layer). Next, an oxide-metal (BZO–Co) nanocomposite layer was grown on top of the insulating layer, showing the VAN growth morphology. The BZO was selected due to the lattice match with the oxide matrix in BZO–Co. This stack configuration allows for the ferromagnetic coupling between the top BZO–Co layer and the lower LSMO which is a semimetal FM phase, thus allowing the electrical tunneling effect across the interface to be controlled by the applied magnetic field (M) as illustrated in Fig. 9(c). Hence, utilizing the VAN framework, the integration of MTJs on Si would be a promising building block for developing future high-density memories and logic circuits using CMOS.

The ME effect, which refers to the coupling effect between electricity and magnetism, has been inspiring the design and development of novel fascinating devices in recent years.^121,122 Owing to the strong ME coupling effect for the electric field control of magnetization or magnetic field control of ferroelectric polarization at room temperature, ME heterostructures open enormous possibilities for the application of novel nanoelectronics devices such as ultralow heat dissipation spin logic memory,^2,123 magnetoelectronic sensors,^124,125 energy harvesting,^126,127 and actuators, etc.^128,129 VAN is definitely an excellent platform for achieving ME coupling by combining FM and FE two phases into one framework. For instance, BFO-CFO grown on STO substrate has shown strong magnetic anisotropy and enhanced ME coefficient, which are particularly useful for designing nonvolatile memories such as magnetoelectric RAM (MERAM) that combines the characteristics of both ferroelectric random-access memory (FRAM) and MRAM with additional functionalities.^130,131 Therefore, more exploration work needs to be done by integrating ME nanocomposite films on the Si substrate for energy-efficient ME switching and memories.

Silicon photonics has been widely recognized as a promising technology for achieving high-density, high-speed, low-cost, and high-performance data processing units.¹³² Over the past decades, much progress has been made in the Si photonic devices such as low-loss waveguide, ultrafast modulator, and high bandwidth detector.^133–135 Hyperbolic metamaterials (HMMs) due to its distinct optical anisotropy and unprecedented optical properties such as negative refractive index,^136,137 subwavelength diffraction^138,139 and superlens,^140,141 have shown wide range of applications such as optical waveguides,^142,143 subsurface sensing,^140,144 and nanoscale resonators, etc.^145,146 Till now, various materials systems including oxide-oxide,^{8–15,20–33} oxide-metal,^{34–41,99–102} nitride-metal,^109–112 and even nitride-oxide¹¹³ and multiple phase complexed VAN systems^{78,103–105,114,147} have been integrated on oxide substrates. However, the Si-integration work for different VAN films is still rare and limited. Therefore, more materials systems need to be explored for integrating VAN films on Si toward complex Si-based integrated photonics.

Although much progress has been made for various VANs integrated on Si, it should be noted that there are indeed some remaining challenges as progress continues to be made toward large-scale Si-integration of VANs for devices. First, creating more regularly shaped and ordered nanostructures is imperative for functional and reliable devices. This could be achieved by using other orientated Si substrates, different buffer layer materials/configurations, and/or various patterning techniques. Most of the works on Si-integration of VANs so far have focused on Si (001) due to its compatibility and domain matching epitaxy with TiN. However, other Si substrate orientations such as Si (111), Si (110), and others are also worth exploring as they could result in unique or improved crystallinity and physical properties for certain VAN systems. On a similar note, different buffer layer materials and/or configurations going beyond the traditional STO/TiN bilayer buffer should also be considered. The VAN systems discussed in this review were primarily based on a cubic crystal lattice structure, but not all VANs are restricted to this structure. For example, hexagonal vertical structures such as those seen in ZnO–Au¹⁰¹ and TaN–Au¹⁰⁹ VAN systems may benefit from a hexagonal lattice buffer material, which could also necessitate a different Si substrate orientation. Thus, it is important to identify and test other buffer layer materials and Si substrate orientations, which are well-suited to the desired VAN thin film. Furthermore, templated techniques such as using miscut-angle or pre-annealed oxide substrates have also been demonstrated to be an effective way for improving the crystallinity and ordering of VAN films, but none of these approaches have been extended to VAN systems integrated on the Si substrate.^102,148 Hence, more endeavors are needed for achieving ordered VAN systems integrated on Si toward reliable device performance.

Another challenge for the large-scale integration of VANs on Si lies with the PLD method used in most of the demonstrations. Although PLD has historically been considered a small area (∼1 cm² or 1 in.²) deposition technique, there has been significant work over the years to achieve wafer-scale PLD.^149,150 In fact, currently, there are several commercially available PLD systems capable of depositing thin films across wafers up to 200 mm in diameter. In addition, it would also be worthwhile to explore the viability of other deposition techniques such as plasma enhanced chemical vapor deposition (PECVD), which is more commonly used for large-scale deposition, to realize VAN thin films on Si. It is apparent that there are numerous possibilities for further research on VANs integrated on Si, including further optimization of the quality and physical properties of the VANs through modification of growth conditions and parameters or even more detailed studies on the interface between the VAN and Si, which could reveal interesting phenomena useful for device applications.

D.Z. and M.K. contributed equally to this work.

This work was supported by the U.S. Department of Energy, Office of Science, Basic Energy Sciences (BES) under Award No. DE-SC0020077. The TEM effort was supported by the U.S. National Science Foundation (Nos. DMR-2016453 and DMR-1565822). D.Z. acknowledges the support from the U.S. Office of Naval Research (No. N00014-20-1-2043).

The data that support the findings of this study are available within the article as well as from the corresponding author upon reasonable request.