The effect of the etching solution concentration on the etching profile of vertical microscale holes formed on a Si(100) substrate by metal-assisted chemical etching (MacEtch) was investigated. MacEtch was performed at different solution concentration ratios (characterized by their molar ratio ρ=[HF]/([HF]+[H2O2])), and the Au catalyst split above a certain concentration ratio. The Au-Si interface in the samples after MacEtch was observed using transmission electron microscopy (TEM), which revealed that there was a porous SiOx interlayer between the Au and Si, and that the SiOx layer became thinner as the concentration ratio increased. The interlayer almost disappeared under solution conditions when the catalyst was split. We propose two hypotheses for the mechanism of catalytic fracture during MacEtch: chemical fracture due to Si atoms diffusing into the grain boundaries of polycrystalline Au by Au-Si interdiffusion, and mechanical fracture due to stress on the catalyst caused by heterogeneous etching.

In this study, we demonstrated formation of high aspect vertical holes in Si substrate using wet chemical approach, MacEtch process, for the application of through silicon via (TSV) holes. It is important that the addition of the surfactants in the etching solution to obtain the well- defined straight holes in the Si substrate using MacEtch, regardless the polarity of the surfactants. MacEtch is expected to be used for formation of TSV instead of Deep reactive ion etching (Deep-RIE) because of lower costs than Deep-RIE [1,2,3]. However, it is difficult to obtain high aspect vertical holes using MacEtch process due to its instability of etching direction for prolonged process as shown in Fig. 1(a). Addition of surfactant is usefulness to improve the morphology of etched holes. Micro scale holes were prepared in N-type Si (100) substrate using MacEtch process. The disc shape Au deposited by sputtering were used as catalyst. Before deposition of the Au catalyst, a Ti interlayer with a thickness of 10 nm was deposited directly on the Si substrate with patterned resist. The patterned Au discs (10 μm in diameter), which were deposited on a Ti / Si substrate using a photolithography lift-off process. The thicknesses of deposited Au were 10 nm for Ti/Si substrate. The base pressure of the deposition chamber was maintained at 10-5 Pa, and purity of Au and Ti target used for the deposition were 99.99 %, respectively. For the MacEtch process, a mixture of 1.0 M hydrofluoric acid (HF), 1.3 M hydrogen peroxide (H2O2) and deionized water was used as the base etching solution. Benzalkonium chloride (BKC), and polyethylene glycol (PEG), which were cationic and non-polar surfactant, respectively, were used as the additive for the etching solutions. The MacEtch of the Si substrate with patterned Au were performed for 120 min at 40 oC in the etching solution. Fig.1 shows cross sectional SEM images of typical shape of etched Si holes using MacEtch. The direction of etched Si hole was changed during MacEtch without surfactants, as shown Fig.1 (a), and these changes of direction occurred more than 60% Si holes. By contrast, we observed drastic improvement on direction of etched holes in almost all holes in addition of PEG, as indicated Fig.1 (b). Similarly, adding BKC in MacEtch process was also observed improvement. The catalyst at the bottom of substrate without surfactant was deformed and partially detached from the Si substrate, as shown Fig.1 (d). However, with PEG and BKC, the catalyst contacted to the bottom and was suppressed deformation, as indicated Fig.1 (e), (f). In addition, BKC suppressed at 25 times low concentration compared with PEG. This phenomenon was is considered due to influence of polarity of hydrogen part or structure of surfactants. Particularly, the shape of catalyst film with BKC kept flat during MacEtch process compared with PEG. Therefore, defamation of catalyst has a strong correlation with formation of bended holes. Moreover, the suppression effect of catalyst is stronger in cation, BKC, than in non-ion, PEG. References [1] L. W. Schaper, S. L. Burkett, S. Spiesshoefer, G. V. Vangara, Z. Rahman, and S. Polamreddy, IEEE Trans. Adv. Packing. 28, 356 (2005). [2] T. Shimizu, R. Niwa, T. Ito, and S. Shingubara, Jpn. J. Appl. Phys., 58, SDDF06 (2019). [3] Y. Asano, K. Matsuo, H. Ito, K. Higuchi, K. Shimokawa, and T. Sato, In Electron. Components Technol. Conf., 2015, p. 853. Figure 1

No abstract

1. Introduction Recently the application of ReRAM to artificial synapse has been paid much attention in the field of neuromorphic computing([1],[2]). For the implementation of neuromorphic synaptic devices, gradual resistance change response by applying voltage pulse trains is required. We investigated the condition of Ti/HfOx/Au-ReRAM to possess gradual resistance change by changing sputtering conditions of HfOx layer. 2. Sample preparation and experiments Figure 1 shows the structure of Ti/HfOx/Au device. For deposition of HfOx film, we used reactive sputtering with Ar and O2 mixture gases. At this time, we adopted two types of Ar/O2 flow ratios, which is (1)7.8 : 15.6 sccm and (2)7.8 : 2.0 sccm. These devices have a bipolar memory characteristic. SET occurs when a positive voltage is applied, and RESET occurs when a negative voltage is applied. We measured resistance change of RESET process of the ReRAM device with applying DC voltage pulses. 3. Results and discussion Figure 2 shows resistance change behavior with applying voltage pulses at RESET process. Pulse width is 1.0 uA and pulse period is 2.0 uA. Figure 2(a) is the resistance change behavior of device (1) with constant amplitude voltage pulses. Resistance change was rarely observed when pulse amplitude is -1.5 V, however, gradual resistance change more than 1 digit was observed when pulse amplitude is -1.6 V. Moreover, when we applied -1.7 V, gradual resistance change more than 2 digits within 40 pulses was observed. In the case of this device, resistance change was ideal for the use of neuromorphic synaptic device. Figure 2(b) is the results of device (2). Gradual resistance change was observed after 60 pulses when pulse width is -1.5 V. However resistance increased more than 3 digits after a few pulses were applied when pulse amplitudes were -1.6 V and -1.7 V. In the case of device (2), resistance change is almost binary and it is not suitable for use of artificial synapse device. The difference of resistance change behavior may come from spatial variation of oxygen vacancies. For further understanding of the mechanisms, we will evaluate the difference of atomic composition of HfOx layer by XPS and XRD.