This latest work may well have solved the problems that have so far inhibited the development of carbon-based memory and opened the door to the possible use of oxygenated amorphous carbon for non-volatile memory applications.
IBM (Zurich) and the Swiss Electron Microscopy Center (Empa) have just published [Ref 1] the details of a new non-volatile memory based on a Redox reaction in thin films of oxygenated amorphous carbon (a-COx) produced by physical vapor deposition (PVD). This latest work reports on the results of device measurements and follows the issuing of an IBM patent US20150036413 A1 earlier this year.
In the past carbon and carbon nanotubes have been shown to have some potential for NV memory applications; processing difficulties, lack of reproducibility and limited write/erase endurance appear to have stifled development in the direction of products. The electrical properties of the other carbon allotrope have been the focus of those pursuing carbon based electronics to challenge silicon or as its follow-on, in either of those roles because of its high electrical resistance amorphous carbon has received less attention.
However, for memory applications the high electrical resistance of amorphous carbon will be of benefit. This latest work may well have solved the problems that have so far inhibited the development of carbon based memory and opened the door to the possible use of oxygenated amorphous carbon for NV memory applications—with the added advantage the ability to use conventional silicon-compatible thin film deposition processes.
The COx structures.
The IBM COxRAM planar memory device employs the familiar planar pore structure favored for many types of thin film memory devices, the central core region of which is illustrated in the upper part of Figure 1.
The memory devices are fabricated on a 500-nm-thick thermal silicon dioxide (SiO2) film formed on a silicon wafer as the substrate. The bottom electrode is a tungsten film, the dimensions of the active area contact to it are delineated by circular pores etched in 35-nm-thick SiO2 film overlaid on the tungsten. With pore diameters ranging from 100nm up to 4 μm. The next step is the physical vapor deposition (PVD) from a graphite carbon target in oxygen of the a-COx active material into the pore to make contact with the bottom electrode. The planar sandwich structure is completed by the deposition of the platinum top electrode metal. There is one very important step before the deposition of the COx, it is the sputter cleaning of the tungsten electrode surface of any native oxide. This will ensure that the part of the Redox action that involves the Tungsten and COx interface is not contaminated or compromised.
Figure 1: (a) In the SET state conducting links of carbon, rings and chains, provide the low resistance paths with a high oxygen concentration near to the W electrode, (b) during RESET to the high resistance state the links are broken and oxygen is released and is moved to saturate the dangling bonds.
The larger diameter structures were for use in the detailed chemical and structural analytical work. A close-coupled series resistor was also integrated into the test structure. Memory devices with titanium and tungsten upper electrodes were also investigated. The majority of the electrical measurements were from W/a-COx/Pt devices with 100 nm diameter pore and nominal thickness of 18 nm.
To bring the memory device to its normal operating state a forming step is required. This consists of applying a triangular shaped pulse of positive polarity to the bottom electrode. When the applied voltage reaches a forming voltage Vf of about 4 to 5 volts, a function of the thickness, the current flowing through the cell abruptly increases, and the cell switches from its virgin state to a low-resistance state (LRS) or SET state. It is also possible to use a sequence of 1μs-wide triangular pulses to form a-COx cells.
Unless forming is indistinguishable from normal operation it represents a problem. A prime example can be found in early phase change memory (PCM) devices where depositing the active material in its crystallized state removed the problem of a high first switching or forming threshold voltage, required if the active material is deposited in its amorphous state. A step which gave PCM its second wind.
I raised the question of the need for forming with Abu Sebastian, the exploratory memory and cognitive technologies researcher at IBM (Zurich), and the possibility of producing a COxRAM device that did not need forming. He said:
“Definitely this is an issue with all carbon-based memories. Franz Kreupl [Ref 2] had a nice write up on this as part of ITRS last year. He calls it the 'first pulse challenge'. My hope is that carbon-memory will scale much better than PCM given its single elemental nature,of carbon it might make this first pulse not that challenging after all.”
Typical voltage and current curves for the SET- RESET operation as a function of time are shown in Figure 2(a)
To bring the device back to a high-resistance state (HRS) or RESET state, a pulse of negative polarity is applied to the bottom electrode without the use of the built-in current limiting resistor. Typical RESET pulses would be of 10 ns duration with amplitude of -4V.
Figure 2:(a) The COx switching characteristics, (b) The symmetrical low field I-V characteristics, first and third quandrant.voltage
For subsequent SET operations, pulses of positive polarity are applied to the bottom electrode with the current limiting Rs in series. Typical SET pulses would be of 50 ns duration with an amplitude of 5V. The minimum pulse width decreases with increasing voltage amplitude to 50 ns at +5V for set pulses and for RESET pulses of -4V of 10ns duration.
While the optimum performance is obtained with bidirectional SET/RESET pulses COxRAM devices can also be operated in the first quadrant of the I-V characteristics. It is suggested that the rupture of the conducting filament which can also be achieved by Joule heating, allowed because the inertness of the Pt electrode prevents any oxidation. It is for this reason other upper electrode materials such as W and Ti offer less write/erase endurance. Filament rupturing would make these new device similar to other carbon-based memories or a-C:H memories.
The IBM team were able to demonstrated write/erase endurance of 5 x 10E4 cycles for a COxRAM device operating with (+5V SET), (-4V RESET) 100ns pulses, maintaining an on-to-off ratio of >10E3 with resistance values for SET/RESET of 10E3Ohms and 10E6 Ohms respectively. For elevated temperature data retention at 85 C, an on-to-off ratio of resistance of >10E3 was maintained for 10E4 secs.