The purpose of this research was to investigate non-volatile memory device technologies that could be applied to reconfigurable electronics applications to provide power reduction, radiation tolerance, smaller size, and improved reliability over existing non-volatile memory devices. The research encompasses: 1) materials and device design, and 2) fabrication and testing of the devices. The types of memory devices that were investigated are divided into three categories:
1) Ion-Conducting, Resistance Variable Memory Devices (a.k.a. memristors). These devices change resistance via the movement of metal ions (Cu and Ag ions) through an active device layer, upon proper application of an electric signal. Devices were designed that showed high speed, low voltage/current operation, variable resistance programmability (through spike-timing-dependent-plasticity tests over timing windows of ns to ms), cycling greater than 1 million cycles, and operating temperatures of at least 150°C without degradation. Many of these materials were tested in two-terminal devices, integrated with CMOS circuits, and in small cross-point arrays.
Recommendation: Based on the factors of device electrical response and ease of fabrication, the ion-conducting resistance variable memory device has the highest potential for successful application to the area of reconfigurable electronics and non-volatile memory. Based on the materials researched, this type of device also has the most flexibility in altering the device electrical characteristics to fit specific applications. Devices can exhibit bi-directional resistance tuning using low voltage, small pulse width signals. The technology can be integrated into a CMOS back-end-of-line process with no consequence to an existing CMOS fabrication facility.
2) Atomic or Molecular Memory Based on Zero-Field Splitting (ZFS). This category comprises devices that define the memory state by the interaction energy of the spin-spin and spin-orbit angular moment of the electrons around an atom and the angular momentum (spin-spin and spin-orbit) of the nucleus. The theoretical operation of this type of device is based on the energy splitting produced by this interaction (in the absence of an externally applied magnetic field), referred to as zero-field splitting. Materials were investigated to try to find a suitable material that exhibited this property at room temperature with enough electron spin density in one of the energy states for a detectable microwave signal absorption upon transition of an electron between energy states. Materials were investigated in bulk form, with some showing promise for this application, as observed via electron paramagnetic resonance spectroscopy. However, deposited thin films of these materials did not show high enough signal absorption to indicate they would be viable for detection at the nanoscale in a memory device. Preliminary device fabrication with these materials also failed to yield a functional ZFS device. This could be due to the electron density being too small in a given state to achieve a detectable signal to noise ratio.
Recommendation: The ZFS device concept is still high risk and theoretical. One approach for exploring the viability of this theory would be to investigate organic molecules as potential candidates since organic molecules frequently exhibit large spin polarization at room temperature. This increase in signal intensity due to spin polarization could give rise to a detectable signal within the small size of a bit.
Phase-Change Memory. Work in this category included devices consisting of stacked chalcogenide (S-, Se- or Te-containing material) thin films and phase-change alloys that are potentially capable of producing multiple memory states. Devices fabricated at Boise State University were large compared to the current technology node (1 um diameter vs < 20 nm). This has significant consequences for device operation due to device volume dependence (for melting and quenching the volume of material to change phase). In addition to the larger two-terminal Boise State devices, other devices fabricated on integrated circuit die with feature sizes at 0.5 um were tested. Phase-change stack materials on smaller-sized devices (40 nm) were also tested using Micron Technology’s 300 mm wafer test process flow.
Recommendation: It is difficult to change the resistance of a phase-change device consistently and between multiple values. The operation of this type of device depends significantly on the device structure and on the materials and fabrication processes. The energy required to change the resistance of a device is significantly higher than the ion-conducting memory device due to the need to heat the volume of material past the melting temperature. The device structures and fabrication processes needed in order to have lower energy switching and consistent device operation are much more complex, which translates into being more prone to processing errors. However, devices that operate with both the phase-change and an ion-conducting mechanisms within the same device material are viable.
This work was done by Kristy A. Campbell of Boise State University for the Air Force Research Laboratory. AFRL-0247
This Brief includes a Technical Support Package (TSP).
Reconfigurable Electronics and Non-Volatile Memory Research
(reference AFRL-0247) is currently available for download from the TSP library.
Don't have an account? Sign up here.