Carbon dot-stabilized magnesium quantum dots for high-performance memristors and driven pulse diagnosis in traditional Chinese medicine

Abstract

With the gradual exploration of memristors in artificial intelligence (AI) applications, the demand for robust stability of memristors is increasing. Previously, researchers proposed adding metal nanoparticles as charge capture materials to the functional layer of the memristor to improve the uniformity of CFs and increase the stability of the device. However, the directly doping metal atoms to the functional layer of the memristor does not prevent atomic aggregation, which reduces the availability of metal atoms. Here, we first loaded single Mg atoms into carbon to form Mg carbon dots (Mg-CDs), and the doping of Mg-CDs to the memristor device (Mg-CDMD) to stabilize device performance of the memristor and then completed. The Mg-CDMD exhibits record-low SET/RESET voltages (0.2 V/−0.3 V), ultrafast switching speeds (8.9 ns), a high resistance ratio (~10⁶), and exceptional endurance (>10¹⁰ cycles). We also propose its application in medical diagnosis for human pulse monitoring based on the different responses of the device to different pulse trains. The pulse conditions are qualitatively translated into electrical conductance based on the analysis of pulse diagnosis in Traditional Chinese Medicine, and a circuit system is constructed to achieve recognition. By correlating conductance patterns with traditional Chinese medicine (TCM) diagnostics, the system successfully discriminates between normal pulses, string-like pulses, and slippery pulses. This artificial intelligence application based on memristors opens up a new research field and is expected to make significant contributions to the further development of Traditional Chinese Medicine.

Graphical abstract

Keywords

memristor / carbon dots / ultrafast switching speeds / human pulse monitoring

1 Introduction

Memristors, as emerging non-volatile memory devices, have garnered significant attention for their potential in neuromorphic computing and artificial intelligence (AI) applications due to their ability to emulate synaptic plasticity [1, 2]. The capability of parallel computing and the synaptic functions of memristor such as the short-term potentiation/depression, synaptic plasticity can help realize the real in-memory sensing and computing functionality [3−5]. However, challenges such as non-uniform device parameters [6], slow switching speeds [7], and limited scalability [8] hinder their practical implementation. The uniformity of conductive filaments (CFs) in the memristor is recognized by researchers as one of the factors that determine the performance of the memristor. To improve the performance of memristors, researchers have proposed an effective method to add metallic nanoparticles, such as Ag, Au, Mo, Ru, Co, and Cu as charge-trapping materials to improve uniformity of CFs, into the functional layer of memristors [9−12]. However, the aggregation of metal atoms occurs during the repeated addition of electric fields, resulting in the low utilization rate of metal atoms and waste of resources. Therefore, it is necessary to develop innovative strategies to stabilize atomic dispersion while leveraging their quantum effects [13−15].

A frequently used approach is to load the appropriate amount of active metal onto various supports with high surface area and desired anchors, such as carbon materials, biomass, and ultrathin nitrogen-doped holey carbon [16−18]. Carbon dots (CDs) have high electrical conductivity, good chemical stability, and unique edge sites. Loading individual metal atoms into the carbon to synthesize metal-carbon quantum dots is an effective method to prevent a single metal atom from aggregating in caustic media [19, 20]. CDs can be combined with metal due to their available intrinsic vacancies, fast electron transfer properties, high stability, and large surface area. The abundant vacancies and functional groups (e.g., −NH₂, −COOH, −OH) in CDs provide favorable binding sites for coordination stabilization of metal-CDs composites [13]. By doping carbon-modified metal quantum dots instead of traditional metal nanoclusters into the functional layer of the memristor, metal atoms can be utilized to the maximum extent.

In this work, taking Mg as an example, we applied Mg carbon dots (Mg-CDs) as quantum dots to the memristor and fabricated a Ag/HfO₂/Mg-CDs/Pt memristor device (Mg-CDMD). The disadvantages of Mg atom aggregation are avoided and Mg-CDs can enhances the local electric field to uniform the parameters of the device and improve the response speed of the Mg-CDMD [21, 22]. Beyond conventional memory applications, we demonstrate the utility of Mg-CDMDs in biomedical sensing, particularly for real-time monitoring of human pulse. By correlating pulse amplitude, duration, and interval with memristor conductance, the device successfully emulates three traditional Chinese medicine (TCM) pulse types — normal, stringy, and slippery pulses — providing a quantitative framework for TCM diagnostics. Furthermore, a pulse pattern analysis system integrating Mg-CDMD arrays with sensor circuits achieves real-time classification of pulse pattern, bridging advanced materials science with clinical medicine. This work not only enhances the stability of memristors by loading metal atoms onto quantum dots, but also creates an interdisciplinary approach to modernize TCM diagnosis [23].

2 Experimental section

In this experiment, the Mg-CDMD was fabricated by drop-coating method and magnetron sputtering. The device was based on a silicon wafer with a 1-μm-thick SiO₂ layer on the surface as the substrate. A 10 nm thick titanium (Ti) metal film was deposited on the SiO₂ layer by magnetron sputtering, and a 100 nm thick Pt layer was deposited by electron beam evaporation equipment as the bottom electrode. An alcohol solution containing Mg-CDs was then applied to the Pt layer by drop coating method and left at room temperature for several hours until dry. Subsequently HfO₂, MoO₃ and ZrO₂ were deposited on the QDs layer by magnetron sputtering. The sputtering power for all three oxides was 80 W. The base pressure of the sputtering chamber was less than 2 × 10⁻⁴ Pa, and a gas flow of 50 SCCM (standard cubic centimetres per minute at standard temperature and pressure) Ar and 25 SCCM O₂ was introduced to ensure a working pressure of 3 Pa. The sputtering times for the three oxides were 30 min, 20 min and 20 min respectively. Next, for 10 min by magnetron sputtering, Ag with a thickness of 50 nm and a diameter of 100 μm was deposited on top as the top electrode.

3 Results and discussion

The preparation of carbon dots was carried out using Scindapsus aureus leaves, which contain the element Mg and are rich in C, N, and so on [Fig. 1(a)]. A mixture of Scindapsus aureus leaves and ethanol was added to a round-bottom flask and condensed and refluxed for 0.5 hours until the green leaves turned white. Then the heating was stopped and the solution was poured out to obtain a carbon dot solution. The solution was spin-concentrated at 60 °C and then freeze-dried to obtain solid carbon dots [19].

Figure 1(b) shows a transmission electron microscopy (TEM) image of the Mg-CDs, indicating that the Mg-CDs have an average diameter of 3 nm. In addition, the energy-dispersive X-ray spectroscopy (EDX) element mapping images (Fig. S1 in the supporting information) show the uniform distribution of C, N, O, and Mg. The Fourier transform infrared (FTIR) spectra [Fig. 1(c)] show stretching vibrations for N−H (3412 cm⁻¹), C−H (2921.6 cm⁻¹), C=N (1641.1 cm⁻¹), and C−O (1086 cm⁻¹) bonds, indicating the presence of numerous carbon-containing and nitrogen-containing groups. Figure 1(d) is the UV absorption spectrum of Mg-CDs, with the position of the absorption peak located at 206 nm and the absorption intensity is 1.32. Fluorescence spectra [Fig. 1(e)] show that light with a wavelength of 420 nm has the highest relative intensity. As shown in Figs. 1f−i, the C 1s analysis reveals the existence of sp²/sp³ carbons (C−C/C=C, 284.7 eV), oxygenated carbons (C−O, 286.3 eV). The N 1s band shows the presence of pyrrolic N. The O 1s band includes peaks at 532.1 and 533.5 eV for C=O and C−O, respectively. The Mg 1s band includes a peak at 1303.6 eV for metallic Mg [24−27].

Figure 2(a) shows the TEM image of Mg-CDMD and the corresponding area of the fast Fourier transform image Figure 2(b) shows the high crystallinity of the Mg-CDs and the interplanar spacing is 0.21 nm, close to the planar spacing of graphitic carbon [24]. Figure 2(c) presents the results of direct current voltage scanning of Mg-CDMD. Current–voltage (I–V) curves provide critical information for evaluating device performance and are among the primary considerations for testing such devices. It is observed that the SET/RESET voltages of Mg-CDMD are approximately 0.09 V and –0.3 V, respectively. Additionally, the supporting information (Fig. S2) contains our statistical results for the SET and RESET voltages as well as the Gaussian fitting curve of Mg-CDMD, where the SET and RESET values are constrained to 0.03–0.27 V and –0.12 to –0.6 V. To verify the universality of the device parameters, three different batches and three different devices of Mg-CDMD were fabricated for DC voltage scanning, and the statistics of SET/RESET voltages and Gaussian fitting to the measured values are shown in Figs. S3 and S4. To demonstrate the applicability of Mg-CDs to other oxides as a functional layer for memristors, two devices (Ag/MoO₃/Mg-CDs/Pt and Ag/ZrO₂/Mg-CDs/Pt) were fabricated and subjected to DC voltage scanning. The statistics of SET/RESET voltages and Gaussian fitting to the measured values are shown in Fig. S5, confirming the applicability of Mg-CDs in memristor devices with different oxides [28].

In addition, the SET and RESET response speeds of Mg-CDMD are shown in Figs. 2(d) and (e). It is obvious that the response speeds of Mg-CDMD (8.9 ns and 17 ns) are lower than those of most devices reported in the literature (Fig. S6) [5, 10, 21, 28, 29]. Superior high switching speed makes it more suitable for analog data processing [5]. The orange waveform in Fig. 2(d) shows a 1 V positive voltage pulse applied to the device (referenced to the orange axis on the left). The purple line represents the device’s response. 8.9 ns after the pulse was applied, the device current increased to 2 mA (referenced to the purple axis on the right), indicating the formation of CFs and the device’s transition to a low-resistance state. This test reflects the device’s response speed to positive voltage. The orange waveform in Fig. 2(e) shows a 1 V negative voltage pulse applied to the device (referenced to the orange axis on the left). The purple line shows the device’s response. 17 ns after the pulse was applied, the device current decreased to 0.1 mA (referenced to the purple axis on the right), indicating the dissolution of the CFs and the device’s transition to a high-resistance state. This test reflects the device’s response speed to negative voltage. The statistics on Mg-CDMD resistance in high and low resistance states exhibit a high ratio of roughly 10⁶ as shown in Fig. 2(f). A high resistance ratio can reduce misreading during operation, enhancing the application of memristors in circuits [30]. Similarly, statistics of high and low resistance for batch-to-batch, device-to-device, and devices with different oxidesare shown in Figs. S7 and S8. The SET/RESET power of Mg-CDMD can reach 1.51 × 10⁻⁸ W and 1.06 × 10⁻⁷ W, respectively [Fig. 2(g)]. In pulse mode, a positive voltage is applied to the device to switch it to a low-resistance state, followed by a negative voltage to switch it to a high-resistance state. This process is called a cycle. The high and low resistance states of Mg-CDMD can be maintained for 10¹⁰ cycles [Fig. 2(h)], indicating that the device exhibits good endurance properties.

Figures S9(a−c) briefly illustrates the conduction mechanism of Mg-CDMDs. It should be noted that the I−V curve of Mg-CDMDs is presented in linear coordinates as shown in Fig. S9(a), while the blue and green parts are marked in detail in Figs. S9(b, c), respectively, in logarithmic coordinates. The typical I–V characteristics of trap-controlled space-charge-limited current (SCLC) contain three parts: (i) the current and voltage show a linear relationship (ohmic contact); (ii) the current is proportional to the square of the voltage (I ∝ V²); (iii) with increasing voltage, the current increases sharply. As can be seen from Figs. S9(a−c), (i) slope 1 of the device is 1.09 in the high resistance state of the positive voltage scanning mode, showing ohmic behavior. (ii) slope 2 is 1.96, which is consistent with the relationship that the current is proportional to the square of the voltage. (iii) slope 3 is 30.49, showing a sudden increase in current. This indicates that the charge transfer behavior conforms to the SCLC mode in the high resistance state [21, 31−34]. In the low resistance state, the slope is close to 1, indicating that it is in a conductive state at this moment, confirming the formation of CFs [35, 36].

Based on the electrical characterization of the device, we further analyzed the conduction mechanism at the microscopic level. Figure 3(a) shows the initial state of the device. When a positive voltage is applied, the quantum dots (QDs) facilitate electron migration, resulting in a significantly enhanced local electric field compared to that in other regions. This intensified electric field promotes the accumulation of Ag⁺ ions near the QDs, substantially increasing the nucleation probability of Ag atoms at these sites [37]. Once Ag nuclei are formed, the strong local electric field drives the highest electrochemical deposition rate of Ag atoms at these nucleation sites [37], thereby facilitating the growth of Ag CFs and the formation of highly conductive pathways through the HfO₂ layer, as illustrated in Figs. 3(b) and (c). Furthermore, the synergistic effect of the high electric field and localized Joule heating accelerates the migration of Ag⁺ ions, enabling fast switching characteristics [38]. Upon removal of the voltage, the slow diffusion of ions at room temperature ensures long-term retention of the resistance state [39]. When a reverse voltage is applied, the Ag CFs undergo electrochemical oxidation and gradually dissolve, as shown in Fig. 3(d). Driven by the reversed electric field, Ag⁺ ions migrate from the filament tips toward the top electrode, leading to filament rupture starting from the QD regions and eventually restoring the device to HRS, as depicted in Fig. 3(e) [28, 40]. During the resistive switching process of the device, the spatial confinement effect of carbon dots on magnesium quantum dots effectively suppresses the aggregation of magnesium quantum dots. This unique structural design not only promotes the uniform spatial distribution of CFs but also enables the controllable regulation of the resistive switching process, thereby significantly enhancing the performance consistency of the device.

In addition, we tested the performance of the device to simulate biological synapses. In biology, the various responses of synapses to different external stimuli are key to obtaining information [41, 42]. Differences in pulse parameters (amplitude, duration, interval), which represent different external stimuli, affect the migration of Ag⁺ which is similar to the migration behavior of ions in biological synapses. In the first case, the pulse duration and interval of the pulse trains were defined as 100 ns and 400 ns, respectively, while the amplitudes of the pulse trains are 0.5 V, 0.6 V, 0.7 V, 0.8 V and −0.5 V, −0.6 V, −0.7 V, −0.8 V, respectively. The influence of amplitude on the device resistance is shown in Figure S9d,e, which indicates that the larger the amplitude, the faster the resistance changes. In the second case, the amplitude and interval of pulse trains were defined as 0.7 V/−0.7 V and 400 ns, respectively, while the variation range of pulse width was 100–400 ns, to investigate the influence of pulse duration on device resistance. As shown in Figs. S9(f, g), the larger the pulse duration, the faster the resistance changes. In the third case, to investigate the influence of pulse interval on the device resistance, the amplitude and pulse duration of the pulse trains were defined as 0.7 V/−0.7 V and 100 ns, respectively, while the variation range of pulse interval was 100–400 ns. As shown in Figs. S9(h, i), the smaller the interval, the more significant the resistance changes.

According to the different responses of Mg-CDMD to various pulse trains, we considered applying the device to human pulse monitoring. In TCM, doctors can diagnose illnesses based on the intensity and frequency of a patient’s pulse. While a normal person’s pulse beats with equal strength and frequency, a patient exhibits a different pulse pattern compared to that of a healthy individual [43−46]. In pulse monitoring experiments, the device receives electrical pulses derived from pulse signals and converts them into conductance to monitor the intensity and frequency of the pulse. Figure 4(a) shows the variation trend of device conductance with pulse signal amplitude [Fig. S10(a)], where conductance increases as the pulse amplitude increases. The pulse train (with a read voltage of 0.1 V) in Fig. 4b shows ten consecutive pulse beats with consistent intensity, while Fig. 4c shows ten consecutive pulse beats with uneven intensity distribution, with a reset voltage applied to the device after each pulse to place it in a high-resistance state. The device conductance values, represented by star symbols in the figures, exhibit consistent behavior when the pulse beats have uniform intensity, but inconsistent behavior when the pulse beats have non-uniform intensity. Figure 4d shows the trend of device conductance with pulse interval [the pulse signal waveform is shown in Fig. S10(b)], where the conductance increases as the pulse interval decreases. The pulse trains (with a read voltage of 0.1 V) in Figs. 4e and 4f show twenty pulse beats with a time interval of \mathrm{\Delta }t. The device is placed in a high-resistance state after each of the two pulses by applying a reset voltage. The device conductance changes slightly when the pulse beats are at the same interval but significantly when the pulse beats have different intervals. Such change in device conductance can help physicians determine whether a patient’s pulse rate falls within the normal range.

At present, with the support of the existing technology, the pulse wave used to analyze the pulse patterns can be obtained through various sensors [47, 48]. The waveforms of pulse pattern correspond to several common pulse conditions in TCM, such as the normal pulse, the string-like pulse, and the slippery pulse (Fig. S11), and each pulse pattern wave consists of the percussion wave, tidal wave and dicrotic wave. The normal pulse, string-like pulse, and slippery pulse are three representative types of pulse patterns, each reflecting different physiological and pathological conditions of the body. The normal pulse has a steady and even rhythm, indicating healthy organ function, and is commonly found in healthy individuals. Traditional Chinese medicine suggests that patients with hypertension have string-like pulse. When a patient is diagnosed with a string-like pulse, further attention can be paid to the symptoms of hypertension, providing doctors and patients with a potential diagnostic clue. The slippery pulse feels smooth and fluid. If this slippery pulse is observed in women, it may indicate pregnancy. If this slippery pulse is caused by illness, it often indicates the presence of phlegm-dampness, food retention, or other conditions like inflammation, fever, indigestion, colds, and influenza [49, 50]. A normal pulse exhibits balanced pulsation, with its three waveform peaks decreasing in intensity. Taking the normal pulse waveform as a reference, the peak of the tidal wave of the string-like pulse is close to that of the percussion wave, and the dicrotic wave has a higher peak. In the waveform diagram of the slippery pulse, the percussion wave has a higher peak, while the tidal wave and the dicrotic wave have lower peaks [51−54]. To further expand the application of the Mg-CDMDs in pulse monitoring, we attempted to qualitatively express the pulse pattern wave as the device conductance, as illustrated in the schematic of Fig. 5(a). Based on the pulse pattern characteristics described above, we applied electrical pulse signals to an array of three Mg-CDMDs, with each signal representing a wave crest and applied to a corresponding device. As shown in Figs. 5(b−d), the electrical pulse signal amplitudes for the normal pulse are 1.4 V, 1.0 V, 0.6 V, with corresponding conductance values (indicated by star symbols) of 4.64 × 10⁻³S, 4.88 × 10⁻⁴S, 1.88 × 10⁻⁴S, respectively. Figure 5e shows the normal pulse waveform, comprising three electrical pulses and the conductance response of Mg-CDMD. For the string-like pulse, the electrical pulse signal amplitudes are 1.4 V, 1.2 V, and 0.8 V, with corresponding conductance values of 4.08 × 10⁻³S, 1.18 × 10⁻³S, and 3.59 × 10⁻⁴S [Figs. 5(f−h)]. Figure 5(i) presents the string-like pulse waveform, including three electrical pulses and the conductance response of Mg-CDMD. For the slippery pulse, the pulse amplitudes are 1.6 V, 0.5 V and 0.4 V, with corresponding conductance values of 6.13 × 10⁻³S, 1.06 × 10⁻⁴S and 9.69 × 10⁻⁵S [Figs. 5(j−l)]. Figure 5(m) displays the slippery pulse waveform, consisting of three electrical pulses and the conductance response of Mg-CDMD. Qualitative analysis of the patient’s pulse pattern based on the three memristor conductance values allows for more accurate disease diagnosis.

Building upon the pulse pattern analysis capabilities of the aforementioned devices, and integrating sensors with circuits, we have developed a pulse pattern monitoring system capable of real-time detection and analysis of pulse signals at the cun, guan, and chi positions. The system employs three highly sensitive thin-film pressure sensors placed at the cun, guan, and chi positions on the wrist to capture pulse vibration signals [Fig. 6(a), Signal Acquisition Section]. The analog signals output by the sensors are amplified and subsequently fed into a memristor-based circuit. The conductance of the memristor dynamically adjusts in response to changes in the pulse signals. And by analyzing these conductance variations, key pulse signal parameters such as pulse amplitude and waveform are extracted [Fig. 6(a), Signal Analysis Section]. The circuit further processes the collected data to classify and identify the pulse patterns, ultimately outputting the current pulse pattern type [Fig. 6(a), Pulse Pattern Recognition Section]. Specifically, the circuit compares the signal intensities at the cun, guan, and chi positions to achieve pulse pattern recognition. For example, if the pulse amplitude at the guan position is significantly higher than those at the cun and chi positions, the pattern is classified as a slippery pulse; conversely, if the amplitudes at the cun and chi positions are stronger, the pattern is identified as a string-like pulse. The system offers high sensitivity and real-time performance, making it suitable for a wide range of applications, including clinical diagnosis in TCM, health monitoring, and TCM education and research. It provides an efficient and precise solution for automated pulse pattern detection.

In the circuit, the signal is processed through a two-stage differential amplifier circuit. The first-stage differential amplifier compares the signals at the cun and guan positions and the cun and chi positions to generate an output. When a normal pulse is detected, the pulse amplitude differences between cun−guan and between cun−chi should be relatively similar. As a result, the voltage amplitudes output by the two differential amplifiers in the first-stage differential amplifier will be comparable. In this case, the second-stage differential amplifier will compare the voltage amplitudes from the two first-stage differential amplifiers and produce an output voltage close to 0 V. for slippery pulse or string-like pulses, the amplitude differences between cun−guan and between cun−chi are pronounced. Thus, the second-stage differential amplifier will output a voltage signal with a certain amplitude. For a slippery pulse, the pulse amplitude difference between cun and guan is noticeably greater than that between cun and chi, resulting in a positive voltage output from the second-stage differential amplifier. Conversely, for a string-like pulse, the pulse amplitude difference between cun and guan is noticeably smaller than that between cun and chi, causing the second-stage differential amplifier to output a negative voltage. By determining the amplitude and polarity of the output voltage, the corresponding pulse pattern can be identified.

Figure 6(b) illustrates typical output results of the circuit during pulse pattern recognition. When the pulse pattern falls within the normal range, the output voltage of the circuit fluctuates around 0 V. If the output voltage exceeds 0.1 V, it may indicate a slippery pulse, with higher voltages suggesting more severe conditions. Conversely, if the output voltage falls below −0.1 V, it may indicate a string-like pulse. Figure 6(c) demonstrates the system’s effectiveness in recognizing five distinct types of pulse patterns, showing that the system can effectively distinguish among various pulse patterns and accurately assess the severity of abnormal pulse patterns. This optimized design further enhances the system’s practicality and reliability, providing robust support for the modernization and scientific development of TCM pulse diagnosis.

4 Conclusions

In summary, the introduction of single Mg atoms into carbon prevents the aggregation of Mg atoms and maximizes their utilization efficiency. Doping carbon-modified metal quantum dots into the functional layer of the memristor effectively enhances the device’s response speed. In addition, based on the conductance changes of the memristor under stimulation by different pulse patterns, the device can be applied in clinical medicine for human pulse monitoring to determine whether the pulse intensity and frequency fall within the normal range. Even more intriguing is that, in accordance with the principles of TCM pulse diagnosis , memristors can convert pulse waveforms into electrical conductance values. When integrated with a circuit for real-time analysis, they enable accurate and timely diagnosis of a patient’s condition, providing a systematic approach for the future development of TCM.

References

[1] M. Chen , G. Dong , X. Li , Z. Gao , H. Feng , F. Wang , G. Li , and X. Li , Influence of MoS₂ quantum dots size on the properties of memristor devices, Optik (Stuttg.) 207, 163776 (2020)

[2] P. Yao , H. Wu , B. Gao , J. Tang , Q. Zhang , W. Zhang , J. J. Yang , and H. Qian , Fully hardware-implemented memristor convolutional neural network, Nature 577(7792), 641 (2020)

[3] C. Tan , H. Wu , M. Zhao , X. Jili , L. Yang , L. Gao , and Z. Wang , Gate‐switchable BST ferroelectric MoS₂ FETs for non-volatile digital memory and analog memristor, Adv. Funct. Mater. 34(39), 2405293 (2024)

[4] C. Tan , H. Wu , Z. Lin , J. Liu , L. Yang , L. Gao , and Z. Wang , Human memory‐inspired MoS₂/BST ferroelectric phototransistor with synchronous sensory, short-term and long-term memories, Adv. Funct. Mater. 35(4), 2414186 (2025)

[5] Z. Fan , H. Fan , C. He , and X. Yan , Implementation of an artificial neuron circuit model based on high speed AlN stacked threshold switching devices, Front. Phys. (Beijing) 20(3), 034201 (2025)

[6] F. Ye , F. Kiani , Y. Huang , and Q. Xia , Diffusive memristors with uniform and tunable relaxation time for spike generation in event-based pattern recognition, Adv. Mater. 35(37), 2204778 (2023)

[7] J. Chen,S. Jia,L. Gao,J. Xu,C. Yang,T. Guo,P. Zhang,J. Chen,J. Wang,Y. Zhao,X. Zhang,Y. Li, A ZTO-based memristor with tunable synaptic plasticity, Colloids Surf. A 689, 133673 (2024)

[8] S. Hamdioui,L. Xie,H. A. Du Nguyen,M. Taouil,K. Bertels,H. Corporaal,H. Jiao,F. Catthoor,D. Wouters,L. Eike, Memristor based computation-in-memory architecture for data-intensive applications, in: 2015 Design, Automation & Test in Europe Conference & Exhibition (DATE), IEEE, 2015; pp 1718–1725

[9] J. H. Yoon , J. H. Han , J. S. Jung , W. Jeon , G. H. Kim , S. J. Song , J. Y. Seok , K. J. Yoon , M. H. Lee , and C. S. Hwang , Highly improved uniformity in the resistive switching parameters of TiO₂ thin films by inserting Ru nanodots, Adv. Mater. 25(14), 1987 (2013)

[10] X. Yan , L. Zhang , Y. Yang , Z. Zhou , J. Zhao , Y. Zhang , Q. Liu , and J. Chen , Highly improved performance in Zr_0.5Hf_0.5O₂ films inserted with graphene oxide quantum dots layer for resistive switching non-volatile memory, J. Mater. Chem. C 5(42), 11046 (2017)

[11] J. E. Kim , S. Hu , J. Y. Kwon , S. Y. Chun , K. Soh , H. Yun , S. H. Baek , S. Nahm , Y. J. Jeong , and J. H. Yoon , Cluster-type conductive path-based selector-less 1R memristor array for spiking neural networks, Nano Energy 140, 110983 (2025)

[12] Y. Yu , M. Xiao , D. Fieser , W. Zhou , and A. Hu , Nanoscale memristor devices: Materials, fabrication, and artificial intelligence, J. Mater. Chem. C 12(11), 3770 (2024)

[13] Y. Liu,N. Chen,W. Li,M. Sun,T. Wu,B. Huang,X. Yong,Q. Zhang,L. Gu,H. Song, Engineering the synergistic effect of carbon dots‐stabilized atomic and subnanometric ruthenium as highly efficient electrocatalysts for robust hydrogen evolution, SmartMat (2021)

[14] H. Hao , M. Wang , Y. Cao , J. He , Y. Yang , C. Zhao , and L. Yan , Boron‐doped engineering for carbon quantum dots-based memristors with controllable memristance stability, Small Methods 8(9), 2301454 (2024)

[15] T. Yu , J. Li , W. Lei , S. Shafe , M. N. Mohtar , N. Jindapetch , P. Dommelen , and Z. Zhao , The resistance switching performance of the memristor improved effectively by inserting carbon quantum dots (CQDs) for digital information processing, Nano Res. 17(9), 8438 (2024)

[16] Q. Q. Yan , D. X. Wu , S. Q. Chu , Z. Q. Chen , Y. Lin , M. X. Chen , J. Zhang , X. J. Wu , and H. W. Liang , Reversing the charge transfer between platinum and sulfur-doped carbon support for electrocatalytic hydrogen evolution, Nat. Commun. 10(1), 4977 (2019)

[17] K. Jiang and H. Wang , Electrocatalysis over graphene-defect-coordinated transition-metal single-atom catalysts, Chem 4(2), 194 (2018)

[18] Y. Ma , T. Yang , H. Zou , W. Zang , Z. Kou , L. Mao , Y. Feng , L. Shen , S. J. Pennycook , L. Duan , X. Li , and J. Wang , Synergizing Mo single atoms and Mo₂C nanoparticles on CNTs synchronizes selectivity and activity of electrocatalytic N₂ reduction to ammonia, Adv. Mater. 32(33), 2002177 (2020)

[19] W. Li , Y. Liu , M. Wu , X. Feng , S. A. Redfern , Y. Shang , X. Yong , T. Feng , K. Wu , Z. Liu , B. Li , Z. Chen , J. S. Tse , S. Lu , and B. Yang , Carbon‐quantum‐dots‐loaded ruthenium nanoparticles as an efficient electrocatalyst for hydrogen production in alkaline media, Adv. Mater. 30(31), 1800676 (2018)

[20] X. Duan , Z. Cao , K. Gao , W. Yan , S. Sun , G. Zhou , Z. Wu , F. Ren , and B. Sun , Memristor‐based neuromorphic chips, Adv. Mater. 36(14), 2310704 (2024)

[21] X. Yan , Y. Pei , H. Chen , J. Zhao , Z. Zhou , H. Wang , L. Zhang , J. Wang , X. Li , C. Qin , G. Wang , Z. Xiao , Q. Zhao , K. Wang , H. Li , D. Ren , Q. Liu , H. Zhou , J. Chen , and P. Zhou , Self‐assembled networked PbS distribution quantum dots for resistive switching and artificial synapse performance boost of memristors, Adv. Mater. 31(7), 1805284 (2019)

[22] K. Wang , J. Chen , and X. Yan , MXene Ti3C2 memristor for neuromorphic behavior and decimal arithmetic operation applications, Nano Energy 79, 105453 (2021)

[23] F. Cai , S. Kumar , T. Van Vaerenbergh , X. Sheng , R. Liu , C. Li , Z. Liu , M. Foltin , S. Yu , Q. Xia , J. J. Yang , R. Beausoleil , W. D. Lu , and J. P. Strachan , Power-efficient combinatorial optimization using intrinsic noise in memristor Hopfield neural networks, Nat. Electron. 3(7), 409 (2020)

[24] S. Lu , L. Sui , J. Liu , S. Zhu , A. Chen , M. Jin , and B. Yang , Near‐infrared photoluminescent polymer–carbon nanodots with two‐photon fluorescence, Adv. Mater. 29(15), 1603443 (2017)

[25] M. Chen , B. Du , Q. Wang , J. Zhang , X. Zhu , Z. Lin , Y. Dong , F. Fu , and T. Yu , Tuning the aggregation of silver nanoparticles with carbon dots for the surface-enhanced Raman scattering application, Carbon 185, 442 (2021)

[26] H. Song , Y. Li , L. Shang , Z. Tang , T. Zhang , and S. Lu , Designed controllable nitrogen-doped carbon-dots-loaded MoP nanoparticles for boosting hydrogen evolution reaction in alkaline medium, Nano Energy 72, 104730 (2020)

[27] L. Ai , Y. Pei , Z. Song , X. Yong , H. Song , G. Liu , M. Nie , G. I. Waterhouse , X. Yan , and S. Lu , Ligand‐triggered self‐assembly of flexible carbon dot nanoribbons for optoelectronic memristor devices and neuromorphic computing, Adv. Sci. (Weinh.) 10(12), 2207688 (2023)

[28] X. Yan , C. Qin , C. Lu , J. Zhao , R. Zhao , D. Ren , Z. Zhou , H. Wang , J. Wang , L. Zhang , X. Li , Y. Pei , G. Wang , Q. Zhao , K. Wang , Z. Xiao , and H. Li , Robust Ag/ZrO₂/WS₂/Pt memristor for neuromorphic computing, ACS Appl. Mater. Interfaces 11(51), 48029 (2019)

[29] X. Yan , Q. Zhao , A. P. Chen , J. Zhao , Z. Zhou , J. Wang , H. Wang , L. Zhang , X. Li , Z. Xiao , K. Wang , C. Qin , G. Wang , Y. Pei , H. Li , D. Ren , J. Chen , and Q. Liu , Vacancy‐induced synaptic behavior in 2D WS₂ nanosheet–based memristor for low‐power neuromorphic computing, Small 15(24), 1901423 (2019)

[30] G. Wang , X. Yan , J. Chen , and D. Ren , Memristors based on the hybrid structure of oxide and boron nitride nanosheets combining memristive and neuromorphic functionalities, physica status solidi (RRL) 14(1), 1900539 (2020)

[31] Y. C. Yang , F. Pan , Q. Liu , M. Liu , and F. Zeng , Fully room-temperature-fabricated nonvolatile resistive memory for ultrafast and high-density memory application, Nano Lett. 9(4), 1636 (2009)

[32] J. M. Beebe , B. Kim , J. W. Gadzuk , C. D. Frisbie , and J. G. Kushmerick , Transition from direct tunneling to field emission in metal-molecule-metal junctions, Phys. Rev. Lett. 97(2), 026801 (2006)

[33] S. Kim , S. Jung , M. H. Kim , S. Cho , and B. G. Park , Resistive switching characteristics of Si₃N₄-based resistive-switching random-access memory cell with tunnel barrier for high density integration and low-power applications, Appl. Phys. Lett. 106(21), 212106 (2015)

[34] I. H. Im,J. H. Baek,S. H. Park,S. Park,S. J. Kim,J. Y. Kim,Y. Kim,Y. J. Lee,K. J. Kwak,H. J. Lee, A low-power filamentary memristor crossbar array enabled via cubic α-phase stabilized mixed-cation lead halide perovskites, J. Mater. Chem. C 13(20), 10310 (2025)

[35] Y. Xing , C. Shi , J. Zhao , W. Qiu , N. Lin , J. Wang , X. B. Yan , W. D. Yu , and X. Y. Liu , Mesoscopic-functionalization of silk fibroin with gold nanoclusters mediated by keratin and bioinspired silk synapse, Small 13(40), 1702390 (2017)

[36] C. Wang , W. He , Y. Tong , Y. Zhang , K. Huang , L. Song , S. Zhong , R. Ganeshkumar , and R. Zhao , Memristive devices with highly repeatable analog states boosted by graphene quantum dots, Small 13(20), 1603435 (2017)

[37] Q. Liu , C. Dou , Y. Wang , S. Long , W. Wang , M. Liu , M. Zhang , and J. Chen , Formation of multiple conductive filaments in the Cu/ZrO₂:Cu/Pt device, Appl. Phys. Lett. 95(2), 023501 (2009)

[38] B. J. Choi , A. B. Chen , X. Yang , and I. W. Chen , Purely electronic switching with high uniformity, resistance tunability, and good retention in Pt‐dispersed SiO₂ thin films for ReRAM, Adv. Mater. 23(33), 3847 (2011)

[39] J. J. Yang , F. Miao , M. D. Pickett , D. A. Ohlberg , D. R. Stewart , C. N. Lau , and R. S. Williams , The mechanism of electroforming of metal oxide memristive switches, Nanotechnology 20(21), 215201 (2009)

[40] J. H. Weng , M. C. Kao , K. H. Chen , and M. Z. Li , Influences of Cu doping on the microstructure, optical and resistance switching properties of zinc oxidethin films, Nanomaterials (Basel) 13(19), 2685 (2023)

[41] M. Yang , X. Zhao , Q. Tang , N. Cui , Z. Wang , Y. Tong , and Y. Liu , Stretchable and conformable synapse memristors for wearable and implantable electronics, Nanoscale 10(38), 18135 (2018)

[42] J. Lan , G. Cao , J. Wang , and X. Yan , Artificial nociceptor based on TiO₂ nanosheet memristor, Sci. China Mater. 64(7), 1703 (2021)

[43] Y. F. Chung , C. S. Hu , C. C. Yeh , and C. H. Luo , How to standardize the pulse-taking method of traditional Chinese medicine pulse diagnosis, Comput. Biol. Med. 43(4), 342 (2013)

[44] R. Guo , Y. Wang , H. Yan , J. Yan , F. Yuan , Z. Xu , G. Liu , and W. Xu , Analysis and recognition of traditional Chinese medicine pulse based on the Hilbert-Huang transform and random forest in patients with coronary heart disease, Evid. Based Complement. Alternat. Med. 2015, 1 (2015)

[45] A. C. Y. Tang , Review of traditional Chinese medicine pulse diagnosis quantification, Complement. Ther. Contemp. Healthc 10, 5044 (2012)

[46] A. C. Tang , J. W. Chung , and T. K. Wong , Digitalizing traditional Chinese medicine pulse diagnosis with artificial neural network, Telemed. J. E Health 18(6), 446 (2012)

[47] K. C. Lan , G. Litscher , and T. H. Hung , Traditional Chinese medicine pulse diagnosis on a smartphone using skin impedance at acupoints: A feasibility study, Sensors (Basel) 20(16), 4618 (2020)

[48] T. Yang,X. Jiang,Y. Zhong,X. Zhao,S. Lin,J. Li,X. Li,J. Xu,Z. Li,H. Zhu, A wearable and highly sensitive graphene strain sensor for precise home-based pulse wave monitoring, ACS Sens. 2(7), 967 (2017)

[49] C. H. Luo , Y. F. Chung , C. C. Yeh , X. C. Si , C. C. Chang , C. S. Hu , and Y. W. Chu , Stringlike pulse quantification study by pulse wave in 3D pulse mapping, J. Altern. Complement. Med. 18(10), 924 (2012)

[50] Y. Wang,M. Li,Y. Shang,N. Li,Y. Zhao,Z. Shang, Analysis and classification of time domain features of pulse manifestation signals between different genders, in: 2019 Chinese Automation Congress (CAC), IEEE, 2019, pp 3981–3984

[51] S. Fu,S. P. Lai, A system for pulse measurement and analysis of Chinese medicine, images of the twenty-first century, in: Proceedings of the Annual International Engineering in Medicine and Biology Society, IEEE, 1989; pp 1695–1696

[52] J. Y. Lee , M. Jang , and S. H. Shin , Study on the depth, rate, shape, and strength of pulse with cardiovascular simulator, Evid. Based Complement. Alternat. Med. 2017, 2867191 (2017)

[53] Z. Zhang,B. Peng,C. H. Luo,C. C. Tai, ANFIS-GA system for three-dimensional pulse image of normal and string-like pulse in Chinese medicine using an improved contour analysis method, Eur. J. Integr. Med. 42, 101301 (2021)

[54] B. Peng , C. H. Luo , N. Sinha , C. C. Tai , X. Xie , and H. Xie , Fourier series analysis for novel spatiotemporal pulse waves: normal, taut, and slippery pulse images, Evid. Based Complement. Alternat. Med. 2019, 5734018 (2019)

RIGHTS & PERMISSIONS

Higher Education Press