記住我
In this article, we demonstrated RS characteristics and some basic synaptic functions in a Cu/HfO2/TiO2 NWA/FTO memristor. We found this device had good switching performance and multi-level memory ability under different compliance current (ICC). The experimental results indicated that the introduction of the HfO2 functional layer can effectively improve the uniformity of the devices. Moreover, some basic synaptic functions were mimicked, including long-term potentiation/depression (LTP/LTD), paired-pulse facilitation (PPF), and spike timing dependent plasticity (STDP).
The fabrication process of the memristive devices based on TiO2 NWA is shown in Fig. 1(a). TiO2 NWAs were prepared by a kind of the well-developed hydrothermal method.1919. B. Liu and E. S. Aydil, J. Am. Chem. Soc. 131, 3985 (2009). https://doi.org/10.1021/ja8078972 First, 36 ml of diluted HCl solution was prepared with 18 ml of de-ionized water and 18 ml of HCl (38%), and then 0.6 ml of titanium (IV) butoxide was added into it. The above solution was transferred into a 50 ml Teflon liner after being stirred for 15 min, and a piece of well-cleaned FTO was immersed in it with the conducting side facing down. The Teflon liner was put in an autoclave and heated to 150 °C for 2 h in an electric oven for the growth of TiO2 NWAs. Subsequently, a HfO2 layer (∼5 nm) was deposited on cleaned TiO2 NWA samples by atomic layer deposition (ALD), and tetrakis (dimethylamido) hafnium (TDMAH) and H2O were used as the Hf and O sources, respectively. Finally, 50 nm-thick Cu was deposited as TE by using thermal evaporation through a shadow mask. The morphologies of TiO2 NW arrays' samples were characterized using a metallographic microscope and a field emission scanning electron microscope. Raman spectra were obtained on a confocal Raman spectrometer using a 532 nm laser. XPS experiments were performed on a Thermo Scientific ESCALAB 250Xi system with Al Kα (1486.6 eV) as the radiation source. A semiconductor analyzer was used to measure electrical characteristics, in which a bias voltage was applied to TE (Cu) while BE (FTO) was grounded.As shown in the Raman spectrum in Fig. S1(a), four peaks at 145, 238, 445, and 610 cm−1 indicate that the TiO2 NWA samples prepared by hydrothermal growth are rutile TiO2.2020. P. Wu, X. Song, S. Si, Z. Ke, L. Cheng, W. Li, X. Xiao, and C. Jiang, Nanotechnol. 29, 184005 (2018). https://doi.org/10.1088/1361-6528/aab014 Figures 1(b) and 1(c) show SEM images of the synthesized HfO2/TiO2 NWA sample. The average height of the nanowires in Fig. 1(c) is ∼500 nm from a tilted cross-sectional view of the NRAs. After the deposition of the HfO2 layer, the top of nanowires became smoother as shown in Figs. S1(b) and S1(c). Figures 1(d)–1(g) show the XPS spectrum of Ti 2p, Hf 4f, and O 1s. The peaks at 458.7 eV and 464.35 eV in Fig. 1(d) corresponded to Ti 2p3/2 and 2p1/2, proving that titanium elements of the TiO2 layer existed in the form of Ti4+.2121. J. Shao, W. Sheng, M. Wang, S. Li, J. Chen, Y. Zhang, and S. Cao, Appl. Catal., B 209, 311 (2017). https://doi.org/10.1016/j.apcatb.2017.03.008 For the HfO2/TiO2 samples, as shown in Fig. 1(e), the peaks at 17.25 and 18.75 eV are in good agreement with the binding energies for the Hf 4f7/2 and Hf 4f5/2 core levels.22,2322. M. Ismail, U. Chand, C. Mahata, J. Nebhen, and S. Kim, J. Mater. Sci. Technol. 96, 94 (2022). https://doi.org/10.1016/j.jmst.2021.04.02523. D. Zanders, E. Ciftyurek, E. Subasi, N. Huster, C. Bock, A. Kostka, D. Rogalla, K. Schierbaum, and A. Devi, ACS Appl. Mater. Interfaces 11, 28407 (2019). https://doi.org/10.1021/acsami.9b07090 The results of the XPS analysis of O 1s in Figs. 1(f) and 1(g) can prove that the oxygen elements in the sample existed in the form of lattice oxygen (529.9 eV) and oxygen vacancies (531.5 eV).2222. M. Ismail, U. Chand, C. Mahata, J. Nebhen, and S. Kim, J. Mater. Sci. Technol. 96, 94 (2022). https://doi.org/10.1016/j.jmst.2021.04.025Figure 2(a) shows the current–voltage (I–V) characteristics of a Cu/HfO2/TiO2 NWA/FTO memristor, and a DC voltage sweep (0 V → +2 V → 0 V → −3 V→ 0 V) was applied to TE. The RS process could be achieved without a forming process, possibly because of the presence of oxygen vacancies in the synthesized HfO2 and TiO2, which can promote the migration of metal ions and the formation of conductive filaments.2424. K. D. Liang, C. H. Huang, C. C. Lai, J. S. Huang, H. W. Tsai, Y. C. Wang, Y. C. Shih, M. T. Chang, S. C. Lo, and Y. L. Chueh, ACS Appl. Mater. Interfaces 6, 16537 (2014). https://doi.org/10.1021/am502741m When the voltage reached about +1 V, the resistance decreased abruptly, indicating that the device was switched from the initial high resistance state (HRS) to the low resistance state (LRS), which is the so-called “set” process. In this process, ICC was set to 1 mA to prevent the device from breaking down. When the negative voltage was applied to TE, the resistance increased gradually, and the device was “reset” back to the HRS. Figure 2(b) shows the retention characteristics of HRS and LRS (ICC = 10 mA). The on/off ratio is close to 103 and mostly unchanged for 104 s, proving great retention characteristics as a potential nonvolatile memristor for Cu/HfO2/TiO2 NWA/FTO devices.To explore the RS mechanism, the performance of devices with different electrode areas was tested. The statistical analysis of the RHRS and RLRS for the devices with three different electrode areas (50, 100, and 130 μm in diameter) was conducted based on the data, which were obtained over five cycles from ten random memory cells for each sample, as shown in Fig. 2(c). The results indicate that the RHRS is inversely proportional to the electrode size, while the RLRS does not show distinct dependence on it. This is consistent with the conductive filaments (CFs) dominated RS mechanism.2525. Y. Qi, Z. Shen, C. Zhao, and C. Z. Zhao, J. Alloys Compd. 822, 153603 (2020). https://doi.org/10.1016/j.jallcom.2019.153603 In addition, multilevel LRS was achieved under different ICC. I–V curves under different ICC of 1, 3, and 10 mA were shown in Fig. 2(d). Those three different LRSs all showed good retentions of 1000 s in Fig. 2(e). Endurances of those resistance states were also explored, each resistance states could switch 25 times stably as shown in Fig. 2(f). These characteristics prove that the device has the potential to realize multi-level storage.For comparison, Cu/TiO2 NWA/FTO devices were prepared. The dependence of the device resistance on the electrode area is similar to the above devices with the HfO2 layer, as shown in Fig. S2, which can be considered that they have the same RS mechanism. As shown in Fig. S3, Cu/HfO2/TiO2 NWA/FTO devices had better cycle-to-cycle uniformity than Cu/TiO2 NWA/FTO devices. To further investigate the device-to-device uniformity, the statistical analysis of the RS parameters (RHRS, RLRS, and Vset) was conducted based on the data, which were obtained from dozens of devices. It can be seen from Figs. 3(a) and 3(b) that the Cu/HfO2/TiO2 NWA/FTO devices had better device-to-device uniformity than Cu/TiO2 NWA/FTO devices. The distribution of Vset, RLRS, and RHRS was more intuitively displayed in Figs. 3(c) and 3(d). The Vset of devices with the HfO2 layer was about 1 V, while the devices without the HfO2 layer were mostly distributed between 0.5 and 1.75 V. Meanwhile, the Cu/HfO2/TiO2 NWA/FTO devices all had an on/off ratio of above 100. It can be seen that the uniformity of the devices was improved by introducing an HfO2 layer. A comparison of some parameters of the Cu/HfO2/TiO2 NWA/FTO device with other research is shown in Table S1. Low operating voltage, on/off ratio over 100, and forming-free and multilevel storage capability are advantages of the device.In order to further analyze the conductive mechanism of the device, the I–V curves in Fig. S4(a) were fitted into double-logarithmic plots. Figure S4(b) shows the fitted plot of the positive region, and it is obvious that the I–V curve of HRS can be divided into three parts. At a low voltage, the slope is roughly 1.04, corresponding to Ohmic conduction (I ∝ V) dominated by thermally generated carriers. As the voltage increases, the slope also increases to 1.71, corresponding to Child's square law (I ∝ V2), and subsequently rose to 6.81 (I ∝ Vn). The change of slope shows that the carrier transport in the HRS follows the space charge limited conduction (SCLC) model.26,2726. X. Yan, J. Zhao, S. Liu, Z. Zhou, Q. Liu, J. Chen, and X. Y. Liu, Adv. Funct. Mater. 28, 1705320 (2018). https://doi.org/10.1002/adfm.20170532027. H. Zhang, X. Ju, K. S. Yew, and D. S. Ang, ACS Appl. Mater. Interfaces 12, 1036 (2020). https://doi.org/10.1021/acsami.9b17026 For LRS, the slope of curve is approximately 1.07, corresponding to Ohmic conduction, and this indicates that conductive pathways (CFs) appear after the SET process. In this case when the active metal is used as TE, the RS mechanism is usually considered to be the electrochemical metallization mechanism (ECM) controlled by redox reactions of the metal electrode.22. K. Sun, J. Chen, and X. Yan, Adv. Funct. Mater. 31, 2006773 (2021). https://doi.org/10.1002/adfm.202006773 When the positive voltage is applied to the Cu TE, an oxidation process will occur (Cu → Cun+ + e−). Because of low electronegativity and small ionic sizes of Cu,28,2928. W. H. Xue, W. Xiao, J. Shang, X. X. Chen, X. J. Zhu, L. Pan, H. W. Tan, W. B. Zhang, Z. H. Ji, G. Liu, X. H. Xu, J. Ding, and R. W. Li, Nanotechnol. 25, 425204 (2014). https://doi.org/10.1088/0957-4484/25/42/42520429. J. Kim, A. I. Inamdar, Y. Jo, H. Woo, S. Cho, S. M. Pawar, H. Kim, and H. Im, ACS Appl. Mater. Interfaces 8, 9499 (2016). https://doi.org/10.1021/acsami.5b11781 Cun+ ions can form quickly and diffuse into the RS layer. On account of low ion mobility in dielectrics like HfO2, those Cun+ ions will be reduced near TE and form CFs toward BE gradually,30,3130. Y. Yang, P. Gao, S. Gaba, T. Chang, X. Pan, and W. Lu, Nat. Commun. 3, 732 (2012). https://doi.org/10.1038/ncomms173731. M. Saadi, P. Gonon, C. Vallée, C. Mannequin, H. Grampeix, E. Jalaguier, F. Jomni, and A. Bsiesy, J. Appl. Phys. 119, 114501 (2016). https://doi.org/10.1063/1.4943776 and then the two electrodes will be connected, switching the device into LRS. After that, when the negative voltage is applied to the top electrode, the filament will rupture because of the influence of Joule thermal effect and an electric field,32,3332. Y. Zhang, G. Q. Mao, X. Zhao, Y. Li, M. Zhang, Z. Wu, W. Wu, H. Sun, Y. Guo, L. Wang, X. Zhang, Q. Liu, H. Lv, K. H. Xue, G. Xu, X. Miao, S. Long, and M. Liu, Nat. Commun. 12, 7232 (2021). https://doi.org/10.1038/s41467-021-27575-z33. P. Zhuang, W. Ma, J. Liu, W. Cai, and W. Lin, Appl. Phys. Lett. 118, 143101 (2021). https://doi.org/10.1063/5.0040902 and then the device returns to HRS. Because the dielectric constant of HfO2 (25) is lower than TiO2 (80), the electric field in the HfO2 layer will be stronger, so CFs are easier to break in the HfO2 layer.3434. Z. Long, C. Zheng, and J. Li, Phys. Lett. A 386, 126995 (2021). https://doi.org/10.1016/j.physleta.2020.126995 Comparatively, the CFs in the TiO2 NWA part are not easy to break. Applying the positive voltage to the TE again will lead the re-formation of CFs at the HfO2 layer and switch the device into the LRS again. In this case, the formation and rupture of CFs are limited to a thin HfO2 layer instead of anywhere in the nanowire, and this can reduce the random diffusion of CFs in the RS layer. As for the phenomenon of multilevel LRS caused by different ICC, the mechanism is still inconclusive. A widely accepted view is that it can be obtained by engineering of the size of CFs.35,3635. Y. Wang, Q. Liu, S. Long, W. Wang, Q. Wang, M. Zhang, S. Zhang, Y. Li, Q. Zuo, J. Yang, and M. Liu, Nanotechnol. 21, 045202 (2010). https://doi.org/10.1088/0957-4484/21/4/04520236. R. Singh, M. Kumar, S. Iqbal, H. Kang, U. Kim, J.-Y. Park, and H. Seo, Adv. Mater. Interfaces 8, 2100664 (2021). https://doi.org/10.1002/admi.202100664 Under high ICC, the CFs will be thicker, thus further reducing the resistance, which is the basis for realizing some synaptic functions in the later part.A biological synapse can be considered as the basic structure of the biological nervous system to transmit information. The potentials from the pre- or postsynaptic neurons will change synaptic connections (or synaptic weight), triggering excitation or inhibition of synapses, and this characteristic is called plasticity. As shown in Fig. 4(a), the three main parts of the memristor can be one-to-one corresponded to the synapse, conductance can represent the synaptic weight, and the RS process under voltage can be seen as a manifestation of synaptic plasticity.26,3726. X. Yan, J. Zhao, S. Liu, Z. Zhou, Q. Liu, J. Chen, and X. Y. Liu, Adv. Funct. Mater. 28, 1705320 (2018). https://doi.org/10.1002/adfm.20170532037. L. Liu, W. Xiong, Y. Liu, K. Chen, Z. Xu, Y. Zhou, J. Han, C. Ye, X. Chen, Z. Song, and M. Zhu, Adv. Electron. Mater. 6, 1901012 (2020). https://doi.org/10.1002/aelm.201901012 A gradually controlled synaptic weight is closely related to the realization of LTP and LTD, and the constant change of conductivity can also be achieved by applying voltage pulse trains for the Cu/HfO2/TiO2 NWA/FTO devices. To simulate the regulation of synaptic weights, 100 positive pulse voltages of 0.8 V (during this period, the conductivity kept increasing corresponding to the LTP status) and then 100 negative pulse voltages of −1 V (during this period, the conductivity kept decreasing corresponding to the LTD status) were applied. Each pulse lasted for 5 ms, and the interval between two pulses is 10 ms. It can be seen in Fig. 4(b) that conductivity gradually increased and then decreased under LTP and LTD pulse; separately, it is generally believed that this process may be related to the gradual change in the diameter of CFs during the application of a pulse.38,3938. Q. Wu, H. Wang, Q. Luo, W. Banerjee, J. Cao, X. Zhang, F. Wu, Q. Liu, L. Li, and M. Liu, Nanoscale 10, 5875 (2018). https://doi.org/10.1039/C8NR00222C39. T. Wang, J. Meng, Z. He, L. Chen, H. Zhu, Q. Sun, S. Ding, and D. W. Zhang, Nanoscale Res. Lett. 14, 102 (2019). https://doi.org/10.1186/s11671-019-2933-y This process reflects long-term changes in the synaptic weight, and it is the basis to realize long-term memory, called long-term synaptic plasticity. As shown in Fig. S5, similar conductivity modulation can be realized on ten different devices, which proves the reliability of the Cu/HfO2/TiO2 NWA/FTO device.Short-term synaptic plasticity refers to fast and reversible changes of the synaptic weight; it is the foundation of fast learning. PPF is a typical form of short-term synaptic plasticity; to put it simply, a pair of identical synaptic spikes with a certain time interval act on the synapse, and the change in the synaptic weight caused by the latter will be stronger than the former.4040. B. Pan and R. S. Zucker, Neuron 62, 539 (2009). https://doi.org/10.1016/j.neuron.2009.03.025 This phenomenon can also be mimicked by Cu/HfO2/TiO2 NWA/FTO devices. Two continuous positive voltage pulses (1 V for 10 ms, and the pulse interval is 20 ms) were applied to TE, and the response current caused by the second pulse was stronger than the first as shown in the inset of Fig. 4(c). The intensity of change was related to the time interval of the paired-pulse (Δt), as summarized in Fig. 4(c), and the relationship between them could be fitted with the following formula:4141. W. J. Chen, C. H. Cheng, P. E. Lin, Y. T. Tseng, T. C. Chang, and J. S. Chen, ACS Appl. Electron. Mater. 1, 2422 (2019). https://doi.org/10.1021/acsaelm.9b00572 PPF%=A2−A1/A1×100%=C1exp−Δt/τ1+C2exp−Δt/τ2,(1)where A1 was the amplitude of the current inspired by the first pulse, A2 represented the current inspired by the second pulse, C1/C2 were the scaling factors, and τ1/τ2 were the time constant for Δt.The duration of synaptic plasticity is mainly related to the frequency of the applied pulse sequence.42–4442. J. Liu, H. Yang, Y. Ji, Z. Ma, K. Chen, X. Zhang, H. Zhang, Y. Sun, X. Huang, and S. Oda, Nanotechnol. 29, 415205 (2018). https://doi.org/10.1088/1361-6528/aad64d43. S. Majumdar, H. Tan, Q. H. Qin, and S. van Dijken, Adv. Electron. Mater. 5, 1800795 (2019). https://doi.org/10.1002/aelm.20180079544. T. Ohno, T. Hasegawa, T. Tsuruoka, K. Terabe, J. K. Gimzewski, and M. Aono, Nat. Mater. 10, 591 (2011). https://doi.org/10.1038/nmat3054 A single positive pulse (1 V, 50 ms) was applied to the device, then the current increased significantly, and returned to the initial state in a short time (less than 5 s), as shown in Fig. 4(d). When the above pulses were applied for ten times with a time interval of 0.8 s, the response current became larger and hold for a long time (more than 50 s), as shown in Fig. 4(e). In fact, this is the basis of the learning mechanism of the nervous system, and the repeated stimulation will convert short-time memory (STM) to long-time memory (LTM).4343. S. Majumdar, H. Tan, Q. H. Qin, and S. van Dijken, Adv. Electron. Mater. 5, 1800795 (2019). https://doi.org/10.1002/aelm.201800795The transmission of information in the nervous system is bidirectional, and intrasynaptic spikes may come from presynaptic neurons or postsynaptic neurons (pre- and postsynaptic spikes). The order and time interval of the arrival of two consecutive opposite spikes are different, and the changes in synaptic weights will also be different, which is a kind of the Hebbian synaptic learning rule called STDP.4545. Y. Li, Y. Zhong, J. Zhang, L. Xu, Q. Wang, H. Sun, H. Tong, X. Cheng, and X. Miao, Sci. Rep. 4, 4906 (2014). https://doi.org/10.1038/srep04906 This could also be simulated by the Cu/HfO2/TiO2 NWA/FTO devices. Two identical voltage pulse sequences were, respectively, applied to two electrodes at a certain time interval, which simulated pre- and postsynaptic spikes, as shown in the inset of Fig. 4(f). When the presynaptic spike arrived earlier than the postsynaptic spike for a period of time (Δt > 0), the synaptic weight increased, and the device switched to LTP. On the contrary (Δt ΔW) was a function of Δt, as shown in Fig. 4(f), which could be fitted as the following equation according to the Hebbian learning rule:3737. L. Liu, W. Xiong, Y. Liu, K. Chen, Z. Xu, Y. Zhou, J. Han, C. Ye, X. Chen, Z. Song, and M. Zhu, Adv. Electron. Mater. 6, 1901012 (2020). https://doi.org/10.1002/aelm.201901012 ΔW=A+exp−Δt/τ+ Δt>0, −A−exp−Δt/τ− Δt<0,(2)where A+/A− were the scaling factors and τ+/τ− were the time constant for the positive and negative Δt, respectively. These results proved the application prospect of the Cu/HfO2/TiO2 NWA/FTO memristor in artificial synapses.In summary, nonvolatile RS characteristics and some basic synaptic functions were achieved using a Cu/HfO2/TiO2 NWA/FTO memristor. The devices showed low switching voltages, good retention, and better uniformity after introducing the HfO2 layer. Multi-level LRS could also be realized by adjusting ICC. Through analyzing the performance of devices with different electrode areas, we believed that the RS characteristics are related to CFs in the RS layer, which may be composed of Cu atoms. The 1D nanowire structure limited the transverse diffusion of the CFs, and the CFs were easier to rupture and re-formation at the HfO2 layer, which improved the stability of device performance. On this basis, some basic synaptic functions, such as PPF, LTP, LTD, and STDP, were simulated. The present work demonstrates that such a Cu/HfO2/TiO2 NWA/FTO device has great potential in nonvolatile memory and neuromorphic computing devices.
See the supplementary material for the Raman spectrum and SEM image of TiO2 NWAs; the I–V characteristics, the double-logarithmic I–V curves, and conductance modulation test of the devices; comparison of device performance.The author thanks the National Natural Science Foundation of China (Nos. 12025503, 12074293, and 12275198) and the Fundamental Research Funds for the Center Universities (Nos. 2042021kf0066 and 2042022kf1181).
Conflict of Interest
The authors have no conflicts to disclose.
Author Contributions
Yudong Sun: Data curation (equal); Formal analysis (equal); Investigation (lead); Writing – original draft (lead). Jing Wang: Formal analysis (supporting); Investigation (supporting). Dong He: Formal analysis (supporting); Methodology (equal). Menghua Yang: Data curation (supporting); Formal analysis (supporting). Changzhong Jiang: Conceptualization (equal); Methodology (equal); Project administration (equal). Wenqing Li: Data curation (equal); Formal analysis (equal); Investigation (equal); Supervision (equal); Writing – review & editing (equal). Xiangheng Xiao: Conceptualization (equal); Funding acquisition (equal); Investigation (equal); Project administration (equal); Resources (equal); Supervision (equal); Writing – review & editing (equal).
The data that support the findings of this study are available from the corresponding author upon reasonable request.
REFERENCES
1. P. A. Merolla, J. V. Arthur, R. Alvarez-Icaza, A. S. Cassidy, J. Sawada, F. Akopyan, B. L. Jackson, N. Imam, C. Guo, Y. Nakamura, B. Brezzo, I. Vo, S. K. Esser, R. Appuswamy, B. Taba, A. Amir, M. D. Flickner, W. P. Risk, R. Manohar, and D. S. Modha, Science 345, 668 (2014). https://doi.org/10.1126/science.1254642, Google ScholarCrossref, ISI2. K. Sun, J. Chen, and X. Yan, Adv. Funct. Mater. 31, 2006773 (2021). https://doi.org/10.1002/adfm.202006773, Google ScholarCrossref3. D. B. Strukov, G. S. Snider, D. R. Stewart, and R. S. Williams, Nature 453, 80 (2008). https://doi.org/10.1038/nature06932, Google ScholarCrossref, ISI4. C. Sung, H. Hwang, and I. K. Yoo, J. Appl. Phys. 124, 151903 (2018). https://doi.org/10.1063/1.5037835, Google ScholarScitation, ISI5. D. Xue, H. Song, X. Zhong, J. Wang, N. Zhao, H. Guo, and P. Cong, J. Alloys Compd. 822, 153552 (2020). https://doi.org/10.1016/j.jallcom.2019.153552, Google ScholarCrossref6. M. Xiao, K. P. Musselman, W. W. Duley, and Y. N. Zhou, ACS Appl. Mater. Interfaces 9, 4808 (2017). https://doi.org/10.1021/acsami.6b14206, Google ScholarCrossref, ISI7. C. Ebenhoch, J. Kalb, J. Lim, T. Seewald, C. Scheu, and L. Schmidt-Mende, ACS Appl. Mater. Interfaces 12, 23363 (2020). https://doi.org/10.1021/acsami.0c05164, Google ScholarCrossref8. C. Ebenhoch and L. Schmidt‐Mende, Adv. Electron. Mater. 7, 2000950 (2021). https://doi.org/10.1002/aelm.202000950, Google ScholarCrossref9. X. Shan, Z. Wang, Y. Lin, T. Zeng, X. Zhao, H. Xu, and Y. Liu, Adv. Electron. Mater. 6, 2000536 (2020). https://doi.org/10.1002/aelm.202000536, Google ScholarCrossref10. G. Milano, S. Porro, I. Valov, and C. Ricciardi, Adv. Electron. Mater. 5, 1800909 (2019). https://doi.org/10.1002/aelm.201800909, Google ScholarCrossref11. V. Fra, M. Beccaria, G. Milano, S. Guastella, S. Bianco, S. Porro, M. Laurenti, S. Stassi, and C. Ricciardi, Nanotechnol. 31, 374001 (2020). https://doi.org/10.1088/1361-6528/ab9920, Google ScholarCrossref12. Y. Wang, W. Wang, C. Zhang, H. Kan, W. Yue, J. Pang, S. Gao, and Y. Li, ACS Appl. Electron. Mater. 4, 3525 (2022). https://doi.org/10.1021/acsaelm.2c00495, Google ScholarCrossref13. Y. Zhang, X. Zhao, X. Ma, Y. Liu, X. Zhou, M. Zhang, G. Xu, and S. Long, Sci. China Mater. 65, 1623 (2022). https://doi.org/10.1007/s40843-021-1956-9, Google ScholarCrossref14. K. J. Gan, W. C. Chang, P. T. Liu, and S. M. Sze, Appl. Phys. Lett. 115, 143501 (2019). https://doi.org/10.1063/1.5116359, Google ScholarScitation, ISI15. Y. Duan, H. Gao, M. Qian, Y. Sun, S. Wu, J. Guo, M. Yang, X. Ma, and Y. Yang, Appl. Phys. Lett. 121, 073502 (2022). https://doi.org/10.1063/5.0102030, Google ScholarScitation16. X. Ding, Y. Feng, P. Huang, L. Liu, and J. Kang, Nanoscale Res. Lett. 14, 157 (2019). https://doi.org/10.1186/s11671-019-2956-4, Google ScholarCrossref17. C. T. Chou, B. Hudec, C. W. Hsu, W. L. Lai, C. C. Chang, and T. H. Hou, Microelectron. Reliab. 55, 2220 (2015). https://doi.org/10.1016/j.microrel.2015.04.002, Google Scholar
留言 (0)