Qianxi Lai

Ionic/Electronic Hybrid Materials Integrated in a Synaptic Transistor with Signal Processing and Learning Functions

2010 WILEY-VCH Verlag Gmb Signal processing, memory, and learning functions are established in the human brain by modifying ionic fluxes in neurons and synapses. Through a synapse, a potential spike signal in a presynaptic neuron can trigger an ionic excitatory postsynaptic current (EPSC) or inhibitory postsynaptic current (IPSC) that temporally lasts for 1–10ms in a postsynaptic neuron. This enables the postsynaptic neuron to collectively process the EPSC or IPSC through 10–10 synapses to establish spatial and temporal correlated functions. The synaptic transmission efficacy can be modified by temporally correlated preand post-synaptic spikes via spike-timing-dependent plasticity (STDP). For example, if a postsynaptic spike is triggered momentarily after a presynaptic spike by a few milliseconds, the synaptic efficacy will be increased, resulting in long-term potentiation (LTP), but if the temporal order is reversed, the synaptic efficacy will be decreased, resulting in long-term depression (LTD). The synaptic efficacy can also be modified with reversed polarities in STDP in different types of synapses. STDP is essential to modify synapses in a neural network for learning and memory functions of the brain. Electronic materials, devices, and circuits have been explored extensively to emulate synapses, but to date they have not been able to match the synaptic functions in the brain. Synaptic transistors with nonvolatile analog memory were fabricated by integrating a charge-storage or ferroelectric materials onto the gate structure of Si metal-oxide-semiconductor (MOS) transistors, but these devices cannot emulate the essential synaptic dynamic functions such as EPSC/IPSC or STDP. Electronic neuromorphic circuits have been designed and fabricated to supply EPSC/IPSC and STDP, but these nonlinear dynamic analog circuits require many transistors and several capacitors to emulate a single synapse. The large capacitor size, complex architecture, and energy consumption of these synaptic circuits limited the number of synapses that could be integrated onto a single chip to about 10–10. The lack of a small, cheap device with the essential synaptic dynamic properties for signal processing, learning, and memory prohibits the circuits from approaching the scale and functions of the human brain that contains 10 synapses. We have designed and fabricated a synaptic transistor based on ionic/electronic hybrid materials by integrating a layer of ionic conductor and a layer of ion-doped conjugated polymer, onto the gate of a Si-based transistor. In analogy to the synapse, a potential spike can trigger ionic fluxes with a temporal lapse of a few milliseconds in the polymer, which in turn spontaneously generates EPSC in the Si layer. Temporally correlated preand post-synaptic spikes can modify ions stored in the polymer, resulting in a nonvolatile strengthening or weakening of the device transmission efficacy with STDP. A single hybrid transistor can replace presently utilized complex and energyconsuming electronic circuits to emulate the synapse for spike signal processing, learning, and memory, which could provide a new pathway to construct neuromorphic circuits approaching the scale and functions of the brain. The synaptic transistor has a Si n-p-n source-channel-drain structure of a conventional MOS transistor, with the Si channel covered by a 3-nm-thick SiO2 insulating layer (Fig. 1a). A 70-nm-thick conjugated polymer layer of poly[2-methoxy-5(20-ethylhexyloxy)-p-phenylene vinylene] (MEH-PPV) and a 70-nm-thick ionic conductive layer of RbAg4I5 were sandwiched between the gate SiO2 insulator and an Al/Ti electrode. To emulate synaptic functions, presynaptic spikes were applied to the transistor gate, and postsynaptic currents, I, were measured from the source. Postsynaptic spikes were also applied to the source. A spike was composed of a 1ms-wide positive voltage pulse with an amplitude Vþ1⁄4 3–5V immediately followed by a 1 ms-wide negative voltage pulse with an amplitude V 1⁄4 3 to 5V (Fig. 1a, Inset). After the spike, the transistor was operated at its rest state under a subthreshold condition by setting the gate voltage Vg1⁄4 0 V. A drain voltage Vd1⁄4 0.1 V was applied continuously. When a presynaptic spike with amplitudes of Vþ/V 1⁄4 4V/ 5V was applied to the transistor gate, the typical I is

Effect of N2 partial pressure on the microstructure and mechanical properties of magnetron sputtered CrNx films

Abstract In this paper, Chromium nitride (CrN x ) thin films were deposited with reactive magnetron sputtering method. EDX, X-ray diffraction and transmission electronic microscopy were employed to characterize their chemical compositions, phases and microstructures. Microhardness and elastic modulus were evaluated using a microhardness tester and the effect of N 2 partial pressure on the composition, phases, microstructure and mechanical properties of CrN x thin films was investigated. The results show that the phase formation of CrN x thin films varies from Cr+Cr 2 N to single-phase Cr 2 N, and then Cr 2 N+CrN to nearly single-phase CrN with increasing N 2 partial pressure. The microhardness values of these films make a distribution ranging from HV 21.4 to 27.1 GPa, and when the atom ratio of Cr:N is 1:2 and 1:1, thin film reaches peak hardness values (HV 27.1 and HV 26.8 GPa, respectively), while the elastic modulus is maximal (350 GPa) when single-phase Cr 2 N films is formed.

Organic nonvolatile memory by dopant-configurable polymer

We report an organic, nonvolatile memory based on dopant concentration-induced conductance changes in a conjugated polymer. Consisting of a polymer poly [2-methoxy-5-(2′-ethylhexyloxy)-p-phenylene vinylene] (MEH-PPV)/ionic conductor (RbAg4I5) bilayer sandwiched between two metal electrodes, the device is electrically switched between its low-conductance “off” state and high-conductance “on” state reversibly and repeatedly with on/off ratios above two orders of magnitude and pulse durations as short as 1μs when a voltage exceeding its threshold values (>+3.5V or <−3.8V) is applied. The conductance change is attributed to the injection/depletion of iodide dopant ions in the MEH-PPV layer by the applied electric field.

Two-step penetration: a reliable method for the measurement of mechanical properties of hard coatings

A two-step penetration method using the microindentation technique is presented to investigate the mechanical properties of hard coatings in this article. This method employs a large-load indentation step, first to exhibit the influences of substrate deformation and coating thickness on the measured hardness values, and then selects a small load, based on the results of the first step to perform the measurement. Since with this small load, the measurement will not be affected by the substrate, the hardness and elastic modulus of the coatings can be obtained precisely. Experiments with TiN coatings deposited on high-speed steel substrate shows that by using this two-step penetration method, the mechanical properties of hard coatings can be measured accurately and reliably.

Analog memory capacitor based on field-configurable ion-doped polymers

A memory capacitor based on a field-configurable ion-doped polymer is reported. The device can be dynamically and reversibly programed to analog capacitances with low-voltage (<5V) pulses. After the device is programed to a specific value, its capacitance remains nonvolatile. The field-configurable capacitance is attributed to the modification of ionic dopant concentrations in the polymer. The memory capacitors might be used for analog memory, nonlinear analog, and neuromorphic circuits.

Configurable Neural Phase Shifter With Spike-Timing-Dependent Plasticity

A configurable neural phase shifter (NPS) is fabricated for spike neuromorphic circuits by integrating an ionic/Si hybrid synaptic transistor with a conventional MOSFET. NPS can mimic a neural network to process temporal-code pulses by shifting their temporal phases by t ranged between 9.0 and 24.5 ms, which falls in the range observed in biological neural networks. t can also be configured by following the synaptic learning rule of spike-timing-dependent plasticity. The device can significantly simplify the structures and reduce the power consumption of neuromorphic circuits to emulate neural networks for spike signal processing, memory, and learning.

Dopant-configurable polymeric materials for electrically switchable devices

We demonstrate that the conductivity of conjugated poly[2-methoxy-5-(2′-ethylhexyloxy)-p-phenylene vinylene] (MEH-PPV) can be configured by changing the dopant concentrations in the polymer electrochemically when an external electric field is applied. Two types of switchable devices were fabricated based on the dopant configurable polymer: (1) devices were composed of a layer of MEH-PPV and a layer of solid inorganic electrolyte (RbAg4I5) sandwiched between two metal electrodes; (2) devices were only composed of a MEH-PPV layer sandwiched between two metal electrodes, but the MEH-PPV layer was doped electrochemically with various anions such as hexafluorophosphate (PF6−) and trifluoromethanesulfonate (CF3SO3−). When a voltage above a threshold value is applied between the two metal electrodes, depending on polarity, the conductance of the devices can be electrically switched between its low-conductance “off” state and its high-conductance “on” state, and the devices can be switched on and off reversibly and repeatedly.