Memory Cell with Threshold Switch (US8642985) by Frederick T. Chen.
A high-density resistive memory traditionally coupled a transistor and a switching resistance in the same cell. While the functionality was easy to predict with this configuration, the density is often compromised by the need to make the transistor large enough so that its channel resistance does not dwarf the switching element resistance, especially during RESET. For this reason, crosspoint memory cells which couple the switching element with a so-called selector device (essentially a generalized diode). The selector device is a two-terminal device which can be one of two categories: 1) passes current in one direction, suppresses current in the other direction; and 2) passes current in both directions, but at low enough voltages, the current is suppressed due to the device nonlinearity.
For a bi-directional resistive memory, the second category would be the requirement to fulfill. A particularly intriguing nonlinear, bidirectional switching device is the threshold switch, also known as the Ovonic Threshold Switch (OTS), named after the late inventor Stanford Ovshinsky. An exemplary I-V curve for such a device is shown in Figure 1. Due to the S-shape, it is also referred to as an S-Shaped Negative Differential Resistance (SSNDR) device.
Figure 1. I-V curve for a threshold switch.
There are two notable characteristics of this switch. First, below the threshold voltage Vth, the current is maintained at low levels, while once the current at Vth is exceed, filamentation occurs and the current is allowed to increase, in principle without limit. In phase-change chalcogenides where this phenomenon has been observed to occur, the material itself can limit the current as it heats up, eventually enabling crystallization. However, for pure threshold switches with no memory capability, this natural material limitation is not present. Hence, although the OTS can function adequately for isolating cells, during the SET operation it lacks the ability to limit current that a transistor would have. Therefore, another device would have to be added in series to limit the current. As proposed in US8642985, this device need not be anything other than a weakly nonlinear resistor. It may be generally considered a voltage-controlled resistor (VCR).
Figure 2. I-V for a combination of TS, VCR and the resistive switching memory element, with the voltages on the individual devices split out individually.
Figure 2 shows what happens when we do a SET or Forming operation with this combination of OTS, VCR and resistive switching (RS) element. Initially ramping up the voltage, all the voltage is on the OTS (bottom I-V). When the voltage reaches Vth on the OTS, it becomes a very low resistance, with the voltage now transferring to the VCR-RS combination. The I-V for this pair (middle I-V) looks like a 1D1R SET or Forming operation. What actually happens at forming or SET is the voltage on the RS element actually drops (top I-V), with the balance of the voltage going to the VCR acting as the actual current limiter. The amount of drop, in turn is dependent on whether the RS has any internal resistance.
The implementation of the TS+VCR+RS can be as simple as stacking them in series in the same stack. Or it can possibly be implemented in a 3D scheme with the VCR formed on the conductor line sidewalls, to be covered by the RS and TS layers.
Frederick (Fred) Chen (Ph.D. Appl. Phys., Cornell, 1996) is a senior researcher at the Industrial Technology Research Institute in Hsinchu, Taiwan, who manages the ReRAM program and staff and is also currently a Deputy Director in the Nanoelectronic Technology Division.