Portland, Ore. - Creating ultratiny, nanoscale systems is often easier than verifying the accuracy of the resulting structures. Indeed, in some instances the structures can actually be lost. Nanoscale techniques produce minute features, but imaging tools are sometimes too crude to spot breaks in them. To the rescue come atomic-force microscopy and now its interactive "can-do" sibling, thermal dip pen nanolithography.
Traditional atomic-force microscopy (AFM) techniques drag a probe with a 100-nanometer tip over nanoscale structures to record a small deflection, thereby producing an image of the surface by mapping its valleys and peaks. By first dipping the AFM tip in liquid metal, a semiconductor or an oxide, dip pen nanolithography (DPN) can directly write 100-nanometer lines-but can't interactively switch between read/write.
Separately, IBM's Millipede project (http://eet.com/showArticle.jhtml?
articleID=18307263) has demonstrated that a resistive thermal element could be added to individually address and heat any of 10,000 parallel AFM tips for direct writing on a polymer by melting little holes.
Putting these ideas together with "meltable" semiconductors, researchers have recently demonstrated the ability to interactively read/write 100-nm bits and hope to someday scan and repair semiconductor masks, and to read, write and verify entire circuits at the nanoscale.
"Our technique allows you to switch between a nanoscale probe and a soldering iron anytime you want," said William King, an assistant professor in Georgia Tech's School of Mechanical Engineering. "For instance, to repair masks you just scan a mask with the tip, and whenever you find a defect, you turn on the soldering iron and fix it. Our technique is vacuum-compatible with semiconductor fabrication, so next we want to directly write an entire nanoscale circuit."
King performed the work with Paul Sheehan, a research chemist at the Naval Research Laboratory in Washington, and Lloyd Whitman, head of the lab's Surface Nanoscience and Sensor Technology Section. King was assisted by his graduate student Tanya Wright.
In traditional dip pen nanolithography, you coat the tip of the atomic-force microscope probe with a molecule of interest, put the tip on the surface and the molecule diffuses off the tip and onto the surface. By moving it at a set rate, a line of specified width can be drawn. Hence, by switching tips among oxide, semiconductor and metal deposition, entire circuits could be written in this way, although the approach has yet to be demonstrated experimentally.
"DPN is a very powerful technique, but the problem is that unless you break contact with the surface, you can't turn off the deposition. Also, to date, unless you change the temperature of everything, you cannot change the deposition rate," said Sheehan.
Typically, DPN users will slowly move a tip to deposit a line, then quickly scan back over the line to make sure it was done properly, thereby slightly contaminating the work with a faint scan line, since it is impossible to turn off deposition.
The only way to avoid contamination is to switch the writing tip with a clean reading tip. Unfortunately, as soon as the tip is lifted off the surface it loses registration. It takes an arduous calibration step to relocate the work just finished with the reading tip.
Both of these problems are solved by the invention of thermal DPN (TDPN), which coats the tip with a solid that melts when a small resistive heater in the tip is turned on.
"With TDPN you turn on the heater to begin deposition and you turn off the heater after you're done, and suddenly your writing tip becomes a reading tip without losing registration," Sheehan said. "So you can then scan over work and immediately verify and fix any problems." Also, he said, "you can vary the rate of deposition by merely adjusting the temperature."
For the demonstration, the researchers obtained an AFM cantilever from IBM Corp.'s Zurich Research Laboratory. A single tip was used by the team here, but for IBM's Millipede "thermally heated cantilevers were fabricated with 10,000 parallel operating tips-so conceivably you could use them with our TDPN technique to directly write a nanoscale circuit on a wafer," said Sheehan.
Using the silicon-on-oxide cantilever, with a tip radius of 100 nm, the researchers demonstrated a heating time of 20 milliseconds and a cooling time of 50 ms. The cantilever temperature approached 700 degrees C in short pulses and, because the resistive heating element can also be used as a temperature sensor, the researchers were able to calibrate the cantilever temperature to within 1 degrees C.
For the demonstration, the researchers used a convenient molecule to coat the AFM tip-namely, octadecylphosphonic acid (OPA)-which has a melting temperature of 99 degrees C. By dipping and evaporating, the researchers were able to lay down two complete monolayers of OPA on the AFM tip.
By toggling the tip temperature between ambient room temperature (for reading) and 122 degrees C, the researchers were able to create square "bits" with a width of 98 nm and a height of 2.5 nm, the height of a single monolayer.
In the experiment, the writing was performed on a thermal insulator-mica-rather than a thermal conductor (like silicon), so it took as long as 2 seconds for the tip to cool off after switching off the heater. But, the researchers maintain, on silicon the write-to-read cycle can run in the 100-kHz range, making it fast enough for manufacturing.
Likewise, IBM has demonstrated AFM tips sharper than 20 nm, and has used these cantilevers to mark bits in a polymer as small as 23 nm. Since the width of writing is determined by the sharpness of the tip in TDPN, rather than the diffusion pattern the molecules follow during deposition, it should be possible, according to the researchers, to use these 20-nm AFM tips to draw 20-nm features.
In conventional DPN, by contrast, "the only way to change the deposition speed is to increase the global temperature of both the tip and substrate," Sheehan said. "Unfortunately, this increases both the deposition rate and the spreading of the diffusion pattern. Therefore, global heating leads to larger patterns presumably having larger halos of contamination."
The researchers demonstrated local control over heating to achieve fast deposition rates and sharp, minimally diffused features. Careful engineering of the cantilever tip and substrate, the researchers claim, should allow TDPN to write features as small as 10 nm.
Likewise, IBM has already demonstrated large arrays of 10,000 parallel heated cantilevers that can write bits on polymers at more than 100 pixels per second. Thus, the researchers extrapolate that reasonable write times are possible while writing wafer-scale nanocircuitry. Also, the multilayer, multicomponent patterns made possible by writing patterns atop one another should enable entire three-dimensional nanostructures to be directly written to wafers.
For the future, the researchers plan to demonstrate that all their predictions and speculation on possible uses of TDPN are valid.
"Our near-term goal is to create a transistor completely using this technique-directly writing both the semiconducting materials and the metallic leads all with TDPN, probably within six months," said Sheehan.
Beyond nanoelectronics, the researchers claim the technique could create bioanalytical arrays for simultaneously testing large numbers of genes, pharmaceuticals or proteins.