PCM data retention and the impact of crystal electrodes (Part 1)
The introduction and continued use of crystallized active material as one electrode in modern phase change memory (PCM) cells offers a number of fabrication benefits. An investigation designed to search for evidence of electro-crystallization effects suggested an unexpected result—that the poor data-retention characteristics of PCM devices at elevated temperatures are not just an unfortunate phenomenon, but are caused by the designs of present-day structures. Furthermore, because crystal growth rates do not scale with lithography, elevated-temperature data retention (ETDR) performance may be dependent on device structure, making untested predictions of PCM ETDR with lithographic scaling unsafe.
The basis of this investigation is the use of a novel approach that establishes a link between ETDR results and parameters associated with the PCM “set” operation (the writing of the memory to its low resistance crystallized data state). This approach offers the possibility of determining whether the “set” pulse for a particular PCM device structure is optimized in terms of the time and the temperatures involved. The “set” time makes a significant contribution to the total PCM write data bandwidth; thus, to be able to reduce that by optimization would represent a useful step forward.
Linking “set” and data retention
Although not usually linked to the PCM “set” operation, the starting points for this investigation are the published accelerated ETDR testing results for PCM devices. In the case of a PCM device, the mechanism that determines an ETDR failure of the reset, or high-resistance state, is fundamentally the same as the mechanism on which the PCM device depends for its operation, i.e. crystallization. That alone is too superficial to be the key to the approach to be described here, however. What was necessary was to find a more detailed structurally dependent feature that would provide a closer link or common factor to both the “set” and ETDR mechanisms.
The initial step of this investigation is that instead of extrapolating the ETDR test results to longer times, we back extrapolate them to shorter times; that is, to those times (on the order of 100 ns) associated with the programming “set” step of a PCM and beyond (see figure 1). A simple explanation of my methodology is illustrated in figure 1; the three data points of a typical ETDR test on a PCM device are shown as red squares. Those data points usually extend over about two decades as a plot of log t = 1/kT, where t is time to failure, T is temperature, and k is the Boltzman constant; they are then forward extrapolated over four or five decades to obtain the time (in years) and operating temperature limits for data retention.
Using a seeded-bridge model
Given the correctness of the model used to provide the close link between ETDR failure results and a fully crystallized “set” device, such a back extrapolation should be able to point to the “set” times and the crystallization temperatures of the materials involved and a possible means of optimizing those variables. In the event that there is there is a large anomaly between the expected and predicted values, it would suggest that another process is involved that either assists or impedes the crystallization process in the PCM device structure in the presence of electric current and field.
Modern PCM device structures that feature one electrode formed of the memory material that remains in its crystallized state introduce a thermal asymmetry that is sometimes enhanced by making one electrode a heater. This makes consideration of the crystal-growth model the special case of “seeded” growth, in which there is always present a massive nucleation site on which crystal growth can start.
For this investigation, I have used what I have termed a seeded-bridge model for the PCM “set” process. In this model, crystal growth for both “set” and ETDR test failures starts on the crystallized active material electrode as the seed and grows to bridge, or reduce, the inter-electrode gap to result in a particular “set” resistance or a designated failure condition, respectively. While growth from the crystal electrode as a seed is not new idea, the term seeded-bridge better describes what occurs during PCM “set” (see figure 2).
The seeded-bridge model makes establishing a close link and a numerical relationship between the fractionally crystallized device of the ETDR test and a fully crystallized “set” device a simple step. I assume the result of an ETDR test failure represents crystal growth that extends into one tenth of the “reset” material; thus, the black line in Figure 1 that represents a fully “set” device is displaced one decade to longer times. Implicit in the accuracy of the seeded-bridge model is a linear relationship between resistance and fractional crystal growth into the inter-electrode space. I am aware that the values of resistance change that are used to define an ETDR test failure do vary between different groups and this will change the parallel position of the fully “set” line (black) to the back extrapolated line (blue) in figure 1 and the later figures from real devices (see part 2) in a proportional manner.
Back extrapolation of the type I propose must end when the a value of 1/kT representing the melting temperature Tm is reached; by definition at Tm crystal growth rate must drop to zero, meaning the lines of t = f (1/kT) for PCM devices must reach a minimum “set” time. The question, for which at this time I do not have an answer, is that at the temperatures involved, does the crystal growth rate go through a maximum that would cause the back-extrapolated line enter a gentle minimum or does it continue with the same activation energy to Tm? For the moment, and until other evidence comes to light, I consider the minimum in “set” time, where seeded crystal growth is at its maximum rate, to be located at the start of a discontinuity to infinite time at Tm(see figure 1).
If we can establish that crystal growth rates go through a minimum at temperatures well below Tm and when translated into times for the seeded-bridge to cross a given electrode gap, the result is values much greater than known set times, we can consider that to be an indication that effects other than temperature (e.g. electro-crystallization) are assisting the crystallization process.
Outside of any consideration of the true shape of the t = f (1/kT) curve, there is an absolute minimum “set” time dictated by the time required to bring the device and its immediate environs up to the required crystallization temperature in a controlled manner, determined quantitatively by the thermal mass of the device and associated thermal time constant.
Pitfalls and alternative models
Before looking at experimental results, we must consider a number of aspects of the proposed back-extrapolation. The first is the mathematical safety of the magnitude of the back extrapolation of over seven decades from three data points spread over two decades that is required to get from the minutes and hours of the accelerated ETDR tests to the PCM “set” times on the order 50 ns to 200 ns. After seven decades of extrapolation, small errors can become large errors. To further complicate matters, there is actually more than one extrapolation involved in obtaining the characteristics of a representative or average PCM device in a population. This is because the three or so data points of the ETDR test and the associated extrapolation to the claimed data retention performance are not those of real devices; instead, they represent what should perhaps be called a probability or confidence device, obtained by extrapolation of log normal-distribution plots. It is the device that will probably appear if a large enough sample is tested. Although for the purpose of the model presented here we can consider it as a real device, the average device can be up to two decades away in time, displacing the line for the fully “set” device by that amount.
There is an additional and even more important consideration that could act to distort any seven-decade back-extrapolation of the type proposed here. It would come about if the devices from which the original data was derived are of a mixed population with two different failure mechanisms in play; for example, if a small part of the population has a fabrication defect that results in what appears to be more rapid crystallization, or even if two different crystallization process are involved, one for a minority of devices and the other for the majority or normal devices. Shih [1] highlighted this two-mechanism failure problem for PCM devices in a paper at IEDM 2008 and suggested that in one case, crystallization was caused by local spontaneous nucleation and growth in the body of the reset material, and in the other, growth from the crystal electrode.
For the normal forward extrapolation of the ETDR test, the tails of the distribution are of extreme importance and for the back extrapolation to be undistorted and valid requires that only one crystallization process can be in play. In terms of the validity of the back-extrapolation and its relation to the “set” pulse for normal devices, it is necessary to establish whether the original ETDR data used as the source material is from the tail bits, the average population, or a mixture of both; and whether the particular device structure has two mechanisms in play. In the event that any ETDR test data points are for outliers, then the displacement in time of the line for a fully “set” normal device will need to be corrected by half the width of the distribution in time.
In the work of Shih, the possibility that element separation caused by electric current or field might have created regions in the reset amorphous material with lower crystallization temperatures was not considered to be an option to account for the results.
Is dual crystallization a fundamental property of the active material composition or, in the context of a device, a combination of the effects of the history of operation, element separation related composition changes, and the electrode surface structure? The closeness of the activation energies of the tail bits and the normal bits in Shih’s work is surprising when two separate processes are involved, nucleation and growth, versus just growth. There are two possible explanations. The first is that at some point near the surface of the hemispherical crystal electrode, the changed amorphous material composition acts to in effect catalyze localized enhanced crystal growth from the spherical crystal electrode, creating a dendrite structure. The second possibility, and in my view more likely, is that on a statistical basis, the reset process leaves a very small crystal nucleating site on the surface of the bottom electrode (BE); not large enough to affect the two-terminal resistance. Because a very small amount of crystal growth from that nucleating site will have a very large effect on the overall resistance of a hemispherical reset region, it will appear to as a failure much earlier than in the case of those devices in which the same growth is occurring from the hemispherical crystal electrode surface. I think one could argue that the reason why a very small crystal nucleating site might be left on the surface of the BE is because the rate of cooling, especially if the BE is a heater, might be just slow enough.
Avoiding unsafe extrapolation
The question as to the safety of extrapolation can perhaps be turned on its head by questioning the safety and accuracy of the forward extrapolation. Ideally, for the forward extrapolation, the data points should be qualified with the sample size. In the context of the requirement for PCM devices of capacity greater than 1 Gb, a sample size that allows high confidence parts per billion (ppb) is mandatory. This avoids the problem of the unsafe extrapolation of the results for a small sample for failures at the level of parts per million or parts per billion; e.g. a hundred devices producing percentage failure rates over three decades extrapolated to parts-per-billion levels.
One recent publication [2] has even used the word “unlimited” to describe the magnitude of their forward extrapolation. Another aspect of the forward extrapolation that is often overlooked is that PCM devices subjected to the ETDR test are usually above the glass transition temperature, not arrived at by quenching; while devices at the temperature to which the forward extrapolation is made are always in the rapidly quenched “reset” state.
It is also important is to consider and eliminate other possible models for “set,” of which there are a number of candidates. We can eliminate two of those because of the seed condition. The first model, involving a slower process, is based on homogenous growth in which incubation-nucleation and growth occur in each region of the temperature gradient. The second model, involving a faster process, is based on a model in which crystal growth starts from nucleation sites on the sidewalls and both electrodes and even remnant crystallites. Those opposing the use of my “seeded-bridge” model must answer the key question: Why can the presence of a massive crystal nucleus be ignored as the dominant starting point for crystal growth?
Because crystal growth rate is an activated function of temperature, the optimum set pulse will be one that maintains the region close to the growing crystal interface at a temperature that results in the highest growth rate, as illustrated in figure 2. A “set” pulse with a trailing edge is best suited to achieve the required result and will represent the optimum pulse as far as minimizing set time. Historically, there is indirect support for the “seeded-bridge” model of PCM “set”—a trailing edge has always been part of any empirically optimized “set” pulse. Why would a trailing edge be necessary, when a square pulse should be able to raise the temperature to the crystallization temperature? There are a number of possible explanations. Threshold switching is an inherent part of the “set” step. If the assumption is made that, post-threshold switching, the switched material is relatively cool, as would be the case for a purely electronic process, then a hotspot at a temperature higher than the crystallization temperature for maximum growth rate must be created. That hotspot must be created to raise the temperature of the extreme regions of the amorphous material in the device structure to the temperature for the highest crystal growth rate. To bring that hotspot back to crystallization temperature requires a trailing edge.
Alternatively, and in accord with the view of this author, the result of threshold switching is a hotspot and although crystal growth can occur in the material around the hotspot, a trailing edge is necessary to complete crystallization process. In both examples above, the hotspot must be at a temperature higher than that required for crystallization because crystals grow up temperature gradients in order to dump the latent heat of crystallization.
Support for the seeded-bridge model
Pertinent to my seeded-bridge model and the special role of the crystal electrode, and as mentioned earlier paragraphs, Shih et al identified two mechanisms for elevated temperature data retention failures. They reported that devices in the bulk of the distribution fail the ETDR test as a result of crystal growth specifically from the crystal electrode, while those few in the tail of the distribution fail from nucleation-growth in the body of the reset material.
Provided that the tail of the distribution is not related to a fundamental problem of crystallization, the results and mechanism involved for the majority of the devices are more important for this analysis. In some cases, the devices in the tail of the distribution are highly likely to be associated with fabrication defects that one hopes can be removed.
Authors from IBM/Macronix [3] in a recently published paper EPCOS 2012 have revisited the subject, and have now clearly recognized the importance of seeded crystal growth in the operation of PCM devices. Acknowledging that measurements on amorphous candidate PCM films outside device structures “greatly overestimate performance,” they explored seeded growth for GST materials and a new so-called golden composition. Consider a methodology proposed as a means, or tool for evaluating new materials outside of the PCM device structure. Initially, a film of as deposited material GST material is crystallized and a number of laser-quenched amorphous dots are written into it (see figure 3). The amorphous dots are then oven baked and the temperatures and times for the dots to re-crystallize are measured by reflection. The data confirms that amorphous films without “seeds” require higher temperatures for crystallization than for amorphous dots with crystallized material at the periphery.
According to these tests, a temperature difference of about 32ºC for approximately the same degree of crystallization and time is reported. This evidence that the presence of crystal seeds reduces the temperature and time for crystal growth is good support for the role of the crystallized electrode as the source of crystal growth. Any models for “set” or ETDR failure that ignore it lack credibility.
Earlier work [4] has shown that for germanium telluride (GeTe) chalcogenide glasses, crystal seeding can lower the crystallization temperature by as much as 100ºC.
Crystal growth does not scale
If the seeded-bridge model proves correct, there is another extremely important consequence in relation to the claimed ability of a PCM device with a given material composition to scale in dimensions and provide the same ETDR performance. Using the seeded-bridge model, isotropic scaling of PCM structures will result in a smaller reset region that will certainly “set” more quickly. The problem is that because crystal growth rates do not scale, that same device will fail its ETDR test more quickly. For example, using my seeded-bridge model, if a device with a 50-nm inter-electrode gap has extrapolated data retention of 10 years at a temperature of 85ºC and is scaled to 25 nm, the data retention will fall to five years (see figure 4).
Conclusions
As I will show in Part 2 of this article, the “seeded-bridge” model does give some correlation with actual results. It therefore follows if that model is accurate, optimistic ETDR characteristics for a given PCM material composition will not be obtained when the device structure with which they were derived is lithographically scaled. ETDR performance is PCM structure dependant. Any claims that link lithographic scaling and ETDR performance for a particular material composition therefore must be accompanied by a detailed model of the “set” process that is different from the seeded-bridge model and must be demonstrated in a scaled device to be credible. The necessary scaled PCM device demonstration is likely to be costly; even more so if it fails. Furthermore, pursuing materials with improved elevated temperature performance may require sacrificing that part of write-data bandwidth that relates to the width of the “set” pulse.
Much more data on seeded crystal growth at temperatures close to the melting temperature is necessary. The crystal growth scaling argument would also suggest that multi-level cell (MLC-PCM) devices are likely to show different ETDR test results for each data level. A move away, and in a historical sense back, to PCM device structures that do not use crystallized GST as one electrode in order to avoid seeded crystal growth might offer a route to improved elevated temperature data retention for many existing and potential PCM materials.
As well as a subject for discussion, I believe this article provides a starting point for the development of new tools for the prediction and optimization of some device parameters. A useful next step will be the development of a COMSOL simulation model that includes “set” current, temperatures, crystal growth rates, and other essential structure-related variables. It might also establish both the correctness or otherwise of the seeded-bridge model and demonstrate the ETDR scaling effect, as well as allowing other crystallization models to be tested.
Part 2 of this article will explore the use of the back-extrapolation methodology on published ETDR data from a number of different sources including Samsung, IBM and SK Hynix.
References, Part 1
[1] Y.H. Shih, et al., “Mechanisms for retention loss in Ge2Sb2Te5 based Phase Change Memory,” Proc. IEDM2008.
[2] S. H. Lee, et al., “Highly Productive PCRAM Technology Platform and Full Chip Operation: Based on 4F2 (84-nm Pitch) Cell Scheme for 1 Gb and Beyond,” Proc. IEDM11-47.
[3] Simone Raoux, et al., “Comparison of data retention measured by static laser testing and in PCRAM devices,” Proc. EPCOS2012.
[4] Robert E Simpson, et al., “Seeded Crystallization and Advanced Phase Change Materials,” Nano Letters 10, 414 (2010), and Applied Physics Letters 100, 021911 (2012).
About the author
Ron Neale is the former editor-in-chief of Electronic Engineering. Also, he is the co-author of "Nonvolatile and reprogrammable, the read-mostly memory is here," by R.G.Neale, D.L.Nelson and Gordon E. Moore, Electronics, pp56-60, Sept. 28, 1970.
If you enjoyed this article, you will like the following ones: don't miss them by subscribing to :
