PORTLAND, Ore. As downsized silicon devices approach nanometer dimensions, single-molecule memory cells challenge conventional wisdom.
Today every micron-size capacitor in a DRAM cell is fastidiously refreshed every millisecond just to guarantee that bleeding electrons don't float the voltage past the boundary between zero and one. If that's the case, how often will individual molecules and single-electron devices need refreshing? Will nanoscale devices retain the same properties of their micron-size brethren? Will such organic molecules even be able to survive the high temperatures of semiconductor fabrication?
Researchers are attempting to answer these questions by carefully characterizing single-molecule devices. While no one is yet claiming that nanometer-size single molecules are ready to replace micron-size DRAM capacitors, some results suggest that it won't be long before molecular-size devices can be fabricated into reliable memories. "I don't think anybody is saying we can get error-free performance from a single-molecule device, but we are finding that collections of them can be integrated into reliable semiconductor memory devices," said Randy Levine, president and chief executive officer of ZettaCore Inc., a Denver startup specializing in molecular memories.
Recently, ZettaCore founding scientist Jonathan Lindsey, who remains a professor at North Carolina State University (Raleigh), released results showing that molecular memories have charge-retention times several orders of
magnitude longer than DRAMs (minutes vs. milliseconds), can withstand extreme temperatures (400C) and can undergo as many as a trillion read-write cycles.
Separately, University of Arkansas professors Huaxiang Fu and Laurent Bellaiche recently reported simulation results that indicate individual nano-scale ferroelectric devices can also be harnessed reliably as semiconductor memories.
"Ferroelectricity is caused by atomic off-center displacements resulting from a delicate balance between short-range covalent and long-range Coulomb interactions. Consequently, many researchers speculated that the effect would disappear at the nanoscale," said Bellaiche. "But our results show that large, robust off-center displacements exist in quantum dots as small as 5 nanometers."
Bulk ferroelectric materials spontaneously form into nanoscale dipoles, enabling them to transduce electricity. However, Bellaiche said, many doubted that such materials would retain their ferroelectric properties at the nanoscale. To find out, Fu and Bellaiche examined barium titanium oxide in 5-nm-diameter nanoparticles (called quantum dots because of the predominance of quantum confinement effects in particles so small).
They found the ferroelectric transducing effect still present at 5 nm, albeit in slightly less efficient form. Instead of forming long chains, Fu and Bellaiche said, the nanoscale ferroelectric forms into small magnetic vortexes. However, by applying a magnetic field, Fu and Bellaiche were able to "unravel" the vortexes and achieve performance comparable to that of bulk ferroelectrics.
Engineers doubting that molecular-size devices can attain the kind of reliability to which chip makers aspire should pay heed to ZettaCore's Lindsey, who together with fellow founding scientist David Bocian, a professor at the University of California, Riverside, and his student assistants Zhiming Liu and Amir Yasseri, tested real molecular devices.
"Engineers have been worried that organic molecules are too fragile to withstand the high temperatures of semiconductor processing and the constant read/write cycling necessary to refresh memories made from them. But I think our results put that question to rest molecular memories can be both durable and practical," said Lindsey.
The researchers proved their point in Bocian's lab by attaching organic porphyrins a disk-shaped molecule similar to chlorophyll onto a silicon wafer. The molecules can store a variable number of electrons, which act as binary ones and zeros.
In the test, an electron was stripped from each porphyrin molecule, thereby using its ionic state to represent digital ones, whereas the neutral molecule with all its electrons present counted as a digital zero. It is also possible to bleed more than one electron from the molecule to represent more than one bit per molecule, an area ZettaCore is actively researching.
The tests revealed that organic molecules used to represent data as charge can reliably retain their memory for many minutes (compared with milliseconds for DRAMs). The researchers were also able to run a trillion read/write cycles to refresh the memories without an error. The group further demonstrated that organic molecular memories can withstand temperatures of up to 400C.
"These results are very encouraging, since they show that not only can organic molecules be reliable enough to act as memory cells, but that we should be able to fit their fabrication into the normal process steps used to make silicon memory chips," said CEO Levine.
Levine disclosed that ZettaCore was nevertheless not relying on individual or even small groups of organic molecules in its prototype molecular memory chips. Instead, he said that ZettaCore and similar nanoscale memory startups will likely use thousands or even hundreds of thousands of molecules in parallel to ensure that single-molecule errors do not affect stored data.
"Even if we end up using 100,000 molecules for each memory cell, they are so small that a memory cell made from them will still be much, much smaller than any DRAM memory cell," said Levine.
Lindsey's and Bocian's research was funded by ZettaCore and the Defense Advanced Research Projects Agency's Moletronics Program.