HP Labs heralds 3D memristor chip

HP Labs provided details of a prototype chip designed by HP researcher Qiangfei Xia that stacked memristor crossbar memory cells on top of a CMOS logic chip at the conference, which was co-sponsored by the University of California, the Semiconductor Industry Association and the National Science Foundation.

"Xia used imprint lithography to add a memristor crossbar on top of a CMOS logic circuit," said HP Labs Fellow Stan Williams, inventor of HP's memristive memory technology. "He has built an integrated hybrid circuit with both transistors and memristors." Williams and HP colleague Greg Snider previously proposed an FPGA in which configuration bits were located above CMOS transistors in a memristor crossbar.

Memristor crossbars include two titanium dioxide layers between two perpendicular arrays of metal lines. One layer of titanium oxide is doped with oxygen vacancies, making it a semiconductor. The adjacent layer is undoped, leaving it in its natural state as an insulator.

When a crossbar junction is addressed by simultaneously applying a voltage to one crossbar line on the top and bottom layers, oxygen vacancies drift from the doped to the undoped layer. This causes it to begin conducting, turning "on" the memory bit. The bit can again be turned "off" by changing the current direction, whereupon oxygen vacancies migrate back into the doped layer.

According to Williams, HP Labs' memristor-based FPGA demonstrates that a CMOS fab can make integrated memristor/transistor circuits in three dimensions.

Also at the symposium, Snider unveiled a design that used memristors in their analog mode as synapses in a neural computing architecture. Memristor crossbars are the only technology that is dense enough to simulate the human brain, Snider claimed, adding that the HP Labs crossbars are ten times denser than synapses in the human cortex. By stacking crossbars on a CMOS logic chip, variable resistance could mimic the learning functions of synapses in neural networks.

HP Labs and Boston University were recently awarded a contract by the Defense Advanced Research Projects Agency to build the first artificial neural network based on memristors.

Also at the conference, Massimiliano Di Ventra of the University of California at San Diego described how memristors can explain biological learning in amoebas. Amoebas learn to change their behavior in a manner that can be explained by an LC circuit and a memristor.

Di Ventra also presented evidence that microscopic memristive elements are present in unicellular as well as multicellular organisms.

- R. Colin Johnson
EE Times