Organic materials and devices have drawn attention for applications in modern electronic devices due to their excellent processability in large-area circuits; possibility for molecular design through chemical synthesis; high mechanical flexibility, comparable to flexible substrates; low processing cost; lower power consumption; good scalability; multiple state property; three-dimensional stacking capability; and large data storage capacity 1 2 3 4 5. Therefore, organic memories have aroused wide interest for electronic devices in new information technology. Organic memory devices can be divided into several device structures as follows: organic capacitors, organic field-effect transistors, organic diodes, and metal/organic semiconductor/metal junctions. Organic memory devices can further be divided based on the charge storage mechanism as follows: ferroelectric, polymer charge trapping, floating-gate storages, and those with nanoparticle use 5.

Organic memory has recently drawn increasing attention. Using a three-layer stacking structure, a tri-layer (polymer/metallic nanoparticle/polymer) structure has demonstrated a bistable memory effect. This simple tri-layer can be used to construct a memory device in a structure consisting of metal/tri-layer/semiconductor layers, the MIS structure. The tri-layer is most important because it is where charge trapping occurs. The memory function of the trilayer structure can be achieved by storing charge in nanoparticles or nanoclusters, which are interposed between the insulating polymer layers. The semiconductor layer is used as an electronic charge source to be injected into the tri-layer to be trapped by the nanoparticles 6 7 8 9. Among the metallic nanoparticles, gold nanoparticles (AuNPs) possess important properties for device fabrication as well as good memory characteristics, such as easy synthesis approach, high work function 10 11 12, good electron-accepting properties 13, and chemical stability 12. A broad work function improves the retention time and speed of the write-erase process 14.

The formation and properties of AuNP thin films embedded in organic memory have been investigated using various methods including typical reduction method followed by spin coating 15 16, Langmuir-Blodgett film deposition 17, dip coating 18, and template-directed assembly method 19. For AuNPs synthesized using the reduction method followed by spin coating to form thin films, AuNPs were normally non-uniformly distributed. Therefore, more AuNPs can be found at the edges of a substrate due to the centrifugal effect, resulting in non-uniform AuNP dispersion that leads to inconsistent memory device properties. AuNPs synthesized using the Langmuir-Blodgett film deposition require complicated processing steps and involve several chemical reactions. Kim et al. 18 used 3-aminopropyl-triethoxysilane to modify a surface in the self-assembly of AuNPs to produce uniform and stable AuNP adsorption on the dielectric surface. Template use allows the controllable formation of AuNPs with specific geometric characteristics. Meanwhile, the template-directed assembly method is promising for obtaining well-distributed AuNPs for a memory device. However, the polymer template has to be removed using O₂ plasma 19.

The current study presents a novel method in forming AuNPs directly on a substrate. The AuNPs were grown on a seeded ZnO template using the low-temperature sacrificial hydrothermal growth technique. The seeded template refers to the nanosize grain-consisting ZnO layer after annealing at a certain temperature. In a typical ZnO nanorod formation, this ZnO seed template contributes to the formation of well-aligned nanorod arrays 20 21. However, in the present study, the ZnO seed template was eliminated or dissolved during the hydrothermal reaction to form AuNPs directly on the substrate (sacrificial process). Without the ZnO seed template, the AuNP formation cannot be tuned. The ZnO seed template dissolves into the hydrothermal reactive bath during the reaction due to the competitive reaction between Au and ZnO formations, which is attributed to the difference in free Gibbs energy (∆G; ∆G_Au = −421.59 kJ/mol and ∆G_ZnO = 16.08 kJ/mol) 22 23. This preceding approach has the advantages of low-cost setup, uniform AuNP distribution, and adjustable AuNP size and allows large-area fabrication. The effect of the annealing temperature of the seeded ZnO template on AuNP properties after the reaction is explained accordingly. The AuNP formation mechanism using this approach is proposed. The memory properties of PMMSQ-embedded AuNPs are explained in detail.

A n-type (100) silicon wafer was cut into small pieces (dimension of 1 × 1 cm) and used as substrates. The silicon substrates were cleaned using a standard RCA cleaning process to remove the organic and inorganic contaminants from the surface. A polymethylsilsequioxane (PMSSQ) layer was deposited at 2,000 rpm for 100 s on the cleaned substrates to achieve 350-nm dielectric layers. Thereafter, the samples were cured in an oven at 160°C for 1 h (Figure  1a). A 200-nm-thick ZnO thin film was then deposited on each substrate using radio-frequency magnetron sputtering at 200 W (Figure  1b).The samples were annealed at different temperatures: 200°C, 300°C, and 400°C, with a ramp rate and soaking time of 5°C/min and 10 min, respectively, to observe the seed layer effects on AuNP formation (Figure  1c). Thereafter, the AuNPs were grown on the ZnO seed template using a sacrificial low-temperature hydrothermal approach. The ZnO-seeded samples were subjected to a hydrothermal reaction in a preheated oven at 80°C for 4 h. The hydrothermal bath contained 0.1 M zinc nitrate tetrahydrate (Zn(NO₃)₂·4H₂O), 0.1 M hexamethylenetetramine (C₆H₁₂N₄), 0.01 M gold(III) chloride trihydrate (AuCl₄.3H₂O), and 10 mL acetic acid. After hydrothermal reaction, the samples were removed, rinsed with deionized water, and then dried. The thin ZnO layer was eliminated due to sacrificial growth (Figure  1d).

A second PMSSQ layer was then spin-coated on top of AuNPs and cured at 160°C for 1 h (Figure  1e). Aluminum (Al) top (1 × 1 mm) and back contacts were deposited using thermal evaporation (Figure  1f).

The surface morphology of the samples was observed using a field emission scanning electron microscope (FESEM; Zeiss Supra™ 35VP, Carl Zeiss AG, Oberkochen, Germany). The presence of chemical elements was analyzed using an energy dispersive X-ray spectrometer. Phase presence was analyzed using the Bruker D8 X-ray diffractometer (XRD; Bruker AXS GmbH, Karlsruhe, Germany). The memory characteristics of the samples were determined using the Keithley Model 4200-SCS semiconductor characterization system (Keithley Instruments Inc., Cleveland, OH, USA).

SEM images of the ZnO seeds with varying annealing temperatures are shown in Figure  2. Before annealing, more large grains were observed (Figure  2a). After annealing, the grain sizes became more uniform with lesser large grains present. The observation proved that the atoms rearranged to form a more stable microstructure during the annealing process. The size of the grains became smaller until annealing at 300°C (Figure  2c). Further increasing the annealing temperature to 400°C caused more large grain presence due to the diffusion of ions (Figure  2d). A similar result was reported by Chu et al. 24. However, for the hydrothermal growth using a seeded template, the seeds must not be too dense to provide a space for growth during the hydrothermal reaction.

The average seed size of each sample was calculated from SEM images using the ImageJ software (NIH, Bethesda, MD, USA). The ZnO seed sizes of the non-annealed sample and those annealed at 200°C, 300°C, and 400°C were 110, 113, 109, and 104 nm, respectively. The ZnO seeds became smaller with increasing annealing temperature. At high temperature, ZnO atoms vibrate strongly at their lattice positions and exchange energy with neighboring atoms when sufficient thermal energy is supplied during the annealing process. Thus, atoms diffuse to achieve the lowest strain energy 25, forming smaller ZnO seed surface grains (Figure  2).

Figure  3a shows the XRD patterns of the ZnO seed layer template annealed at different temperatures, with a preferential growth along (002) at 34.4° in all samples. The (002) peak intensities increased with increasing annealing temperature, indicating a crystallinity improvement in the ZnO seeds. This crystallinity improvement is due to the sufficient thermal energy of the annealing process, which caused Zn and O atoms to rearrange in a proper site. Thus, the ZnO seed layer template crystallinity is improved 26. Moreover, the (002) peak positions shifted to a higher angle with increased annealing temperature due to compressive stress relief. During the annealing process, the ZnO film atoms gained energy to rearrange; thus, the ZnO seed film achieved relaxation 24 27 28. Meanwhile, ZnO peaks disappear; however, a dominant peak of (111) at 38.3°C appeared, which corresponds to the AuNP presence. Theoretically, for a face-centered cubic metal Au, the surface energies of the low-index crystallographic facets usually increase in the order of {111} < {100} < {110} due to an increase in interatomic distance. The {110} plane is the most favorable plane for the Au atom deposition because {110} has the highest facet surface energy. Therefore, {110} planes have the highest rate of dissolution and recrystallization of Au atoms, eventually resulting in {110} facet disappearance as the Au particles grow, thereby making the synthesis of Au nanoparticle with {110} facets difficult. Therefore, Au prefers to grow along the {111} facets because the {111} facets have the lowest surface energy, which is more stable, as shown in Figure  3b 29 30. The ZnO peaks for all samples were unobserved, confirming the dissolution of the ZnO seed template during the hydrothermal reaction. This dissolution is attributed to the competitive growth between Au and ZnO. Furthermore, ZnO erodes in an acidic solution 31. Therefore, in the present study, the pH of the hydrothermal precursor was similar for all samples at approximately 2.7. Figure  3c shows the cross-sectional structure of the fabricated metal-insulator-semiconductor (MIS) memory device annealed at 400°C. The SEM micrograph showed that AuNPs were embedded in between the two PMSSQ layers.

Figure  4 shows the morphology of AuNPs grown on the ZnO-seeded substrates annealed at various temperatures after the hydrothermal reaction. The AuNPs for the sample grown on the non-annealed template were spherical and non-uniformly distributed. Furthermore, the AuNPs tend to agglomerate. Fewer large-sized AuNPs were observed for the sample grown on the template annealed at 200°C. Few AuNPs were grown isolated, but some agglomeration was observed. A greater amount of AuNPs with a smaller diameter than the first two samples was obtained for the sample grown on the ZnO template annealed at 300°C, indicating that the AuNP area density was high. More AuNPs were isolated compared with the first two samples. A higher AuNP amount for the samples grown on a template annealed at 400°C was produced, and AuNPs were closer to each other, with some attached to one another.

The AuNP size and distribution were affected by the ZnO seed size. Figure  2 shows that the ZnO seed size decreased with increasing annealing temperature. Thus, more sites were available for AuNP nucleation. With the same Au ion amount, small Au nanoparticles can be formed. The distance between the AuNPs for large seeds was far. Some large AuNPs were also observed due to AuNP interdiffusion.

The average AuNP diameters for the samples grown at different annealing temperatures of the ZnO template were calculated using the ImageJ software. The average AuNP diameters for the non-annealed samples and those annealed at 200°C, 300°C, and 400°C were 113, 120, 90, and 70 nm, respectively. The AuNP number per unit area (area density) was calculated. The area densities of the non-annealed sample and samples annealed at 200°C, 300°C, and 400°C were 2.6 × 10¹³, 8.9 × 10¹², 2.7 × 10¹³, and 4.5 × 10¹³ m⁻², respectively.

Theoretically, a template determines the size, shape, and distribution of nanoparticles 32. This basic principle was applicable in this study, wherein the AuNP size and distribution were confined by the template. AuNPs nucleate on a high surface energy site, which is the ZnO grain boundary. The grain boundary area between the ZnO seeds increased and lead to the formation of larger AuNPs when the ZnO seed size increased. The AuNPs grown were larger because of the larger ZnO seeds. Therefore, low AuNP area density was observed.

AuNPs were successfully grown directly on PMSSQ using a sacrificial templated growth hydrothermal reaction at a low temperature. The AuNP size and distribution were dependent on the annealing temperature of the ZnO sacrificial template because AuNPs were nucleated and grown on ZnO seed grain boundaries. The AuNP size decreased, whereas the area density increased, with increasing ZnO annealing temperature. Optimum memory properties of the PMMSQ-embedded AuNPs were obtained for AuNPs grown on the ZnO seed template annealed at 300°C, with an estimated 54 electrons per Au nanoparticle and excellent I-V responses between an erased and programmed device.

The authors declare that they have no competing interests.

LPG and NSR performed the experimental works and drafted the manuscript. KAR designed the experimental works as well as the writing of the manuscript. CKY and KCA participated in designing and advising the electrical properties analysis. PCO performed the electrical properties analysis and data interpretation. All authors read and approved the final manuscript.