Demonstration of a RRAM-CMOS random number generator circuit for flexible hybrid electronics

Abstract A hybrid random number generator (RNG) circuit is designed by utilizing stochastic switching variation, an inherent property arising from the randomness in formation and dissolution of the conductive filament of a flexible resistive random access memory (RRAM) device. Initially, a widely accepted simple one-dimensional (1D) RRAM compact model is calibrated with the experimental data from a fabricated flexible RRAM device, with less than 5% of root mean square error. Then, two model parameters reflecting the cyclic variations are identified and incorporated into the compact model using simple mathematical functions. Moreover, the effect of random variation of these two physical parameters, (a) the initial gap length of the conductive filament from the opposite electrode and (b) the activation energy, on the switching characteristics of the device is investigated. Furthermore, a simple four-bit RNG circuit is proposed that utilizes the switching variations of the RRAM device as a source of entropy or randomness. A complementary metal–oxide–semiconductor-based circuit is designed using XOR, AND, and four-bit counter blocks to produce a four-bit random number utilizing a 65 nm UMC . The output of this flexible hybrid circuit is tested for randomness by utilizing the benchmark test suite NIST SP 800-22, and it is found that the output passes most of the applicable tests without incorporating any post-processing scheme, indicating the presence of randomness. These results indicate that the incorporation of variation in a simple 1D model can be qualitatively achieved with the above-discussed parameters, and a simple flexible hybrid RNG circuit may be implemented for hardware security applications.