A team from IBM, Zurich have moved to threshold switching as the next part of their focus on understanding every detail of the conduction mechanisms associated with nano-metric sized phase change memory (PCM) devices.
Manuel Le Gallo and a team from IBM, Zurich [Ref 1] have moved to threshold switching as the next part of their focus on understanding every detail of the conduction mechanisms associated with nano-metric sized phase change memory (PCM) devices [Ref 2 EETimes]. Although their conclusions that threshold switching is essentially a thermal effect do not yet quite merit shock and horror headlines, the results are sure to send out a few shock waves to the advocates of purely electronic threshold switching mechanisms for amorphous materials chalcogenides and oxides.
This new result should not come as too great a surprise. In a recent paper K D Tsendin [Ref 3] at the Ioffe Physical-Technical Institute, Russian Academy of Sciences, St.Petersburg, accounted for eleven different electronic mechanisms which in the past 50 years have been proposed to account for threshold switching in amorphous materials and the very rapid transition between the insulating and conducting states. If any had been correct then they would have surely become the accepted and developed as the norm.
As will be discussed in later paragraphs these new results might provide a reason why some of these electronic theories exist and also the lack of high frequency, or any, oscillators based on using the negative resistance transition in chalcogenides. It is to be hoped the thermal design implications of this new work will not be lost on those attempting to construct 3D crosspoint memory arrays by stacking multiple threshold switching devices.
The IBM attack on the problem was divided into a number of parts, summarized in Figure 1, consisting of 1(a) the measurement of the power at threshold switching, 1(b) I-V characteristics, 1(c) delay time measurements and 1(d) simulation and modelling. For the nanoscale sized PCM devices with high thermal efficiency and small thermal time constant were the experimental devices the resulting thermal dynamics leads them to the conclusion that a thermally-assisted switching mechanism is the only plausible interpretation of their results.
Figure 1: Establishing the case for thermal switching (a) constant power at the threshold voltage,(b) the hysteresis effect (c) delay time and (d) simulation and modelling.
Click here for larger image
Using a field and temperature dependent sub-threshold conductivity model [Ref 2] coupled with thermal feedback, quantitative agreement with experimental switching data was obtained for realistic values of the equivalent thermal resistance and capacitance. They were able to accurately reproduce threshold switching dynamics without introducing any electronic threshold switching mechanism, the switching coming solely from the Joule-heating induced thermal feedback combined with high electric field Poole-Frankel conduction.
This raises the question of the possibility that Poole-Frenkel conduction alone, the dominant conduction mechanism prior to threshold switching can provide the positive feedback mechanism required for threshold switching and “S” shaped negative resistance. In the past it has been proposed that tunnelling from deep traps could lead to threshold switching.
The IBM, Zurich team's response to that is if PF is the only mechanism involved the threshold voltage would increase with increasing temperature, which is the exact opposite of what is observed experimentally. This is because the tunnelling current would need to be higher than the thermally activated current to produce an instability that can lead to switching.
The term “high thermal efficiency” is important and does allow for the possibility of other mechanisms where large low efficiency device structures are involved. Where high thermal efficiency is a high value of thermal resistance with low value of thermal capacity, i.e. a nanometric volume.
One of the key pieces of new evidence for the thermal assisted explanation is shown in Figure 1(a). where threshold current is plotted as a function of threshold voltage for measurements over the temperature range 40 C to 160 C and bracketed by two curves of constant power, at 15µW and 22µW.
The IBM evidence for a thermistor-like behaviour or what they describe as a thermal hysteresis effect is shown in Figure 1(b) and shows no evidence of a separate conducting state. This thermal hysteresis effect was obtained by applying a triangular shaped pulse with a leading and trailing edges of 50ns duration. Here the device was driven to its threshold switching point and then before it completes its full transient excursion the current was reduced. It is an excursion into a conducting state of hot amorphous material without any evidence of a conductivity discontinuity or a transition to a different conducting state. These measured results are accurately reproduced by a model based on thermal dynamics and Poole-Frenkel conduction. Consideration must be given to the possibility that this might be an unusual situation because the current is reducing from the moment the maximum voltage is reached and threshold switching occurs.
Pre-threshold switching delay time measurements is another piece of supporting evidence. It is well known that for PCMs the application of an over-voltage above Vt, the steady state switching threshold voltage, there is a delay time prior to switching. With the delay time inversely proportional to the over-voltage. The IBM team looked at this switching incubation time in detail. It was characterized by an increase in current just prior to switching indicative of pre-heating prior to the rapid transition. As shown in Figure 1(c) a value of current reaching 40µA was used as the delay time marker.
For “dome” PCM structures the steady state simulation produced an interesting result it was the formation of hotspots or a hot torus at the edge of the bottom electrode as illustrated in Figure 1(b). Although often ignored by others this would be expected from geometric considerations if as expected the “dome” structure was perfectly centred on the bottom electrode.