Kovio Inkjet Prints Fast Silicon Transistor

David Lammers, News Editor -- Semiconductor International, 11/14/2007

In a claimed breakthrough for printed electronics, Kovio Inc. (Sunnyvale, Calif.) today announced that it has printed an nFET thin-film transistor (TFT) with electron mobilities of 80 cm²/Vs.

Kovio printed an nFET thin film transistor (TFT) with electron mobilities of 80 cm2/Vs.

The company used ink jet printing to deposit electronic inks containing doped silicon, insulators, metal and other materials required for the multilayer device, which used 20 µm design rules.

Kovio CEO Amir Mashkoori

Amir Mashkoori, CEO of the 31-person company, said Kovio is already making integrated CMOS devices in its lab, and the company’s roadmap includes creating RFID tags with &1000 transistors by the end of next year at cost targets far below what conventional subtractive silicon processing can achieve. The company's RFID tag chips would start out at the 20 µm design rules, though he said 10 µm design rules have been used in the company labs.

Kovio uses stainless steel substrates similar to those used in photovoltaics. Longer term, the company seeks to apply its printing technology to solar and display applications, although the initial target is RFID tags costing just pennies each.

Mashkoori was set to describe the advance at today’s keynote speech at the IDTechEx Printed Electronics USA 2007 show, being held in San Francisco this week.

“What we are announcing is the world’s first all-printed high-performance silicon transistor,” Mashkoori said in an interview. 

In an April 2006 article in Nature, researchers working at Seiko Epson Corp. (Fujimi, Japan) announced an ink jet-printed polycrystalline silicon transistor with a mobility of ~6 cm²/Vs, which exceeded the capabilities of amorphous silicon transistors by about six times. Much of the industry's research has been in relatively slow organic transistors, based on pentacene and other plastics, rather than in silicon.

Mashkoori said its high-performance TFT required a combination of electronic inks with the required viscosity and purity, the printing process, and the device architecture. “All those things have to come together to make proper devices out of printed electronics, and it is very difficult to get that combination,” he said.

Nanocrystals of electronic materials are used to formulate electronic inks.

Kovio claims that with its relatively high mobilities, relatively low threshold voltages, and CMOS architectures, it can create CMOS RFID tags with low-power consumption. The transistors are fast enough to support the standard International Standards Organization (ISO) interfaces required to communicate with RFID readers and other parts of the RFID ecosystem.

Kovio was co-founded by Joe Jacobson of the MIT Media Lab, and includes University of California Berkeley professor Vivek Subramanian as its key advisor. The company  has been in stealth mode for six years, and recently achieved $20M in additional venture funding, enough to move to commercialization, Mashkoori said.   

“The concept of silicon ink been out there for 30 years. We believe this ability to get to these mobilities is a key building block to making things very, very cheaply, which is not easy to do with conventional silicon processing. Through an additive process using printing, we can save substantional amounts of money,” he said.

While conventional fabs can cost billions and are adept at putting as many as a billion transistors on a single die, the goal with printed electronics is to create manufacturing facilities that are in the $10M range. These fabs use a fraction of the chemicals required in subtractive processing, and electrical power consumption is also far less.

Kovio VP Vikram Pavate

Electronic inks developed by Kovio and its partners are now achieving sufficient purity and performance, said Vik Pavate, vice president of business development. And electronic printers, including Gravure, laser jet and ink jet types, are increasing the amount of material that can be deposited during each second of processing, a key parameter to achieving high volumes at low costs.

Mashkoori said there are many applications, such as the billions of bottles and cans made each year, that would like to embed some form of intelligence and communications capabilities, providing the costs are low enough.

“Lot of applications want to be cheaper, but it is very difficult to make things very cheaply” with conventional silicon processing, he said.

Reaching the goal of printing a few thousand transistors per chip very cheaply requires an ability to deposit high-performance silicon inks quickly. Rather than focusing on the number of transistors per square centimeter, a key parameter in printed electronics is “how much material you can lay down in one place per second. Once you get the performance capability that we have now developed, it is all about how to do it faster. Our goal is to deposit more material very, very cheaply, instead of how small can you get the piece of silicon,” Mashkoori said.

If RFID tags can be made for a few pennies each, they will be cheap enough to embed in packaged consumer goods costing a dollar or two at unit volumes that could reach into the trillions.

“Intel just announced 45 nm processors, a billion-transistor product. We are talking about less than a thousand transistors, but in unit volumes of about a trillion. Our goal is more about item-level intelligence instead of concentrated intelligence,” Mashkoori said.

Pavate said that in 1970, Intel introduced the 4004 processor using 10 µm design rules to integrate about 2200 transistors. Four decades later, printed electronics is likely to be using similar design rules and transistor counts to make RFID tags costing pennies each. The result, he said, is a form of reverse Moore’s Law scaling, but at process costs that are far lower than conventional silicon subtractive processing.