From the Capital Equipment pages of Electronic News: April 27, 1998 Issue
Materials
FeRAM Must Pay Its Dues
Promise of future is current payoff as many firms spend the R&D bucks
By Chad Fasca
With more than three decades and billions upon billions of dollars poured into fine-tuning silicon-based semiconductor manufacturing, "Trying a new thing is not as easy as it sounds," says Peter Kirlin, executive VP of Advanced Technology Materials, Inc. (ATMI). Consequently, promising materials like ferroelectrics must pay their dues before entering the fabs and subsequent mainstream devices.
At the moment, ferroelectric RAM (FeRAM) technology is paying its dues with promise. Industry sources say that NEC, Fujitsu, Toshiba, Mitsubishi, Siemens, Hyundai, Matsushita and Motorola all have R&D programs involved in ferroelectric memory.
"They want to invent the next-generation DRAM," says Joe Raynak, VP of sales and marketing at Novtek Test Systems, a ferroelectric material test equipment supplier. "Eventually, I think (FeRAM) will work its way into a DRAM architecture." However, he believes it will happen in steps. And those of the most interested parties agree with this assessment.
In FeRAM, Colorado Springs, Colo.-based Ramtron International Corp. has been clearing a path to production for years. Formed in 1984, Ramtron has been working on making memory devices based on novel materials, particularly ferroelectrics.
"We have the arrows in our back to show that we are pioneers," says Ramtron CEO L. David Sikes.
Ramtron has had FeRAM products on the market since 1992, although the products have been directed at low-level designs.
"Even in low density, there is a billion-dollar market in which we think this material can foot the bill," says Mr. Sikes. He noted that a significant market is in EEPROM and battery-backed SRAM.
According to Mr. Sikes, no one questions the advantages of the material or its potential usefulness. Ferroelectric materials include multicomponent oxides such as lead zirconate titanate (PZT), strontium bismuth tantalate (SBT) or Ramtron competitor Symetrix Corp.'s Y-1 material. The key feature of these materials is that they are non-volatile. Unlike typical volatile memory used in PCs, non-volatile memory retains data even when power is shut off, therefore eliminating the need to be constantly refreshed in order to hold information.
With a low-cost, high-yield form of ferroelectric memory, the RAM, ROM and EEPROM on a smartcard could all be replaced. Memory takes up an estimated 80 percent of a smartcard. Ferroelectric memory has been known to function in any one of the other memory capacities, but its cost versus yield in implementing has been the major drawback. Consequently, ferroelectric memory has gotten a bad rap, according to Mr. Sikes. The word is that it can't be manufactured. It doesn't last. It doesn't operate at low voltage. And that it can't be scaled down.
However, he claims that Ramtron's partnership with Rohm is proving that FRAM can be manufactured. In the area of degradation and low voltage, he feels that recent developments in Ramtron's Advanced Materials group have addressed these concerns.
The Advanced Materials group disclosed on April 15 that it has developed a new high-endurance ferroelectric capacitor structure based on an optimized PZT materials composition. According to Ramtron, testing of the new capacitor structure shows minimal switch charge degradation after an equivalent of 1015 read/write cycles, which the company claims is an endurance improvement of five orders of magnitude over the company's current PZT capacitor structures. Further, at 1015 cycles, the structure exhibits a switch charge margin factor of three to four times what is required to change memory logic states, according to Ramtron.
"What this lab development has shown is that we can build things that will last forever," says Mr. Sikes.
In addition, at the recent International Symposium on Integrated Ferroelectrics, the company announced it has successfully scaled its PZT-based process stack for 3-volt and 2V operation. The scaled, low-voltage, PZT stack exhibits a switch charge margin factor of four and also produces excellent aging and low leakage characteristics.
"The last remaining issue is: Can we scale this to megabit-, gigabit-, terabit-type products?" says Mr. Sikes.
The real challenge to building these FeRAM products is integrating the technology into standard semiconductor flows. "That is what we have spent 10 years and $105 million on," says Mr. Sikes.
Ramtron does not expect to tackle this hurdle on their own. They have partnerships with Hitachi, Rohm, Toshiba, Fujitsu, Samsung, SGS-Thomson and Ulvac, among a few others. "Everybody's got a small little project going," he says.
FeRAM technology has been moving closer to this promise. At the Institute of Electrical and Electronics Engineers' International Electron Devices Meeting, NEC and Fujitsu Laboratories disclosed new integration technologies for memory cell architectures which have FeRAM equaling or shrinking DRAM cell architecture. Toshiba also has demonstrated a way to fabricate FeRAM cells much smaller than their DRAM counterparts, according to a recent article in Japan's Nihon Kezai Shimbon.
According to an NEC Research spokesperson, by improving the fabrication process, overcoming the degradation issue, NEC expects to bring FeRAM products to market after 2000, according to the spokesperson. The research spokesperson did say that NEC's work in the area of integration technologies is not material specific, instead the developments are about memory cell operation. The company has compared materials from several suppliers.
"It is really an interface issue, not a materials issue." He notes that Ramtron has found a number of materials that exhibit ferroelectric capabilities and could prove useful to product developments.
"The real key is managing the way you form the interface and the composition of the capacitor plates that are used for the building of the capacitor structure. It is really more of an engineer's job, than of engineering materials work," he says.
"What is really important is the building of new products based on this technology. We could care less, whether it is PZT, Y1 or peanut butter. The product is what is important from our perspective."
According to Mr. Sikes, Ramtron has received production quality shipments from Rohm, and has ordered 1.5 million units deliverable in the first half of this year.
Ramtron last week said that U.K.-based Ampy Automation placed a follow-on order for 500,000, 4-kilobit FRAM (or FeRAM) memory products. Ampy will use the FRAM devices in their growing line of electronic residential power meters. In addition to Ampy, Ramtron is shipping FRAM memory to major electronic power meter suppliers in Europe, Africa and Asia.
The Smartcard IC Interest
------------------------------------------------------------------------
Another company in the FeRAM hunt is ATMI. In February of 1997, ATMI completed a four-year effort to develop the use of barium strontium titanate (BST) for the capacitors of future DRAMs, specifically those of the 256-megabit and one-gigabit generations (EN, Feb. 24, 1997). ATMI was the prime contractor under a Defense Advanced Research Projects Agency-sponsored consortium, with IBM, Micron Technology and Texas Instruments--the three largest U.S. manufacturers of DRAMs--and Varian Associates, the Palo Alto, Calif.-based semiconductor equipment supplier; Aachen University of Technology and North Carolina State University.
Results of the program exceeded expectations, ATMI said. The properties of the BST capacitor offer the potential to significantly reduce the complexity of 1G DRAMs, and overcome several barriers to integrating ferroelectric materials into mainstream semiconductor manufacturing.
To capitalize on these efforts, late last year ATMI formed a joint development program with TI to develop high-density, non-volatile ferroelectric memory technology for smartcard ICs (EN, Nov. 17, 1997). The company also established a new division, Emosyn, as a fabless semiconductor start-up to compete against the likes of Motorola, Siemens, SGS-Thomson and Hitachi in the pursuit of the smartcard IC market. Motorola announced its own joint venture for FeRAM with Japan's Matsushita.
"My sense is we are two years away, even under an optimistic scenario, from having a FeRAM part that can take away significant market share, whether it is a smartcard IC or EEPROM," says Dr. Kirlin. He qualified his optimistic scenario by saying that yields would have to improve dramatically over the next six to 12 months to ensure such a timetable.
"The promise is undoubtedly there," says Dr. Kirlin. "That is what is drawing all the investment."
The integration issues are big, so you need a big market to drive the manufacturing requirements and then the others can follow with little additional investment, according to Dr. Kirlin.
"At one gig, BST will hit manufacturing first, all the manufacturing issues will be solved because that is a $21 billion market, and then PZT or Symetrix SBT or Y1," will follow BST into DRAM manufacturing, he says. BST is a ferroelectric material that has been detuned by doping for its high dielectric constant.
Emosyn, a division of ATMI, has set its sights on GSM handsets where power-conscious SIMM modules present a significant opportunity for ferroelectrics. The potential reduction in power consumption would extend battery life in the GSM handsets. A secondary target is ferroelectric memories in smart card ICs to replace magnetic swipe cards for electronic point-of-sales transactions. However, until then, Emosyn will bide its time, while its investigative efforts with TI continue, hoping to bear ferroelectric fruit.
"You work very diligently on solving the problems, but you don't bet your life on it," says Dr. Kirlin. "Our first products will not use ferroelectrics, they will use standard IC technology you can buy off the shelf, and novel features that will allow us to compete." Dr. Kirlin notes that Emosyn will have two product generations preceding implementation ferroelectrics on its roadmap. Ferroelectrics are slated to enter their products in early 2002. Emosyn has placed FeRAM on a longer leadtime partly because the company sees several years before FeRAM is in production, but also to insulate itself from any obstacles in bringing it to market.
ATMI and TI will investigate the appropriateness of chemical vapor deposition lead zirconium titanate films for possible implementation in the conventional, industry standard one-transistor, one-capacitor (1T1C) cell architecture appropriate for high-density/small-cell-size applications--one of the more important developments of the DRAM consortium.
Ancillary Voices
------------------------------------------------------------------------
Despite being years away from volume production of DRAM architecture devices, FeRAM has caught the interest of some providers of ancillary services, most notably test services.
"Ferroelectric material is a mixture of chemicals. It is the structure that makes the material unique. We make a tester that studies the character of the ferroelectric materials. We are studying it really at the material phase and not at the memory phase," says Joe Raynak of Novtek Test Systems.
Novtek recently revealed at the International Symposium of Integrated Ferroelectrics an agreement with Symetrix Corp. to license their Othello ferroelectric material test software and instrumentation.
"Once the ferroelectric memory is made into a device, there is nothing special about the tester, but there is something to testing the material, for testing the chemistry," says Mr. Raynak.
Under its contract with Symetrix, Novtek will manufacture, distribute and support Othello in exchange for worldwide distribution rights. The two companies also agreed to cooperate on product enhancements and new generations of ferroelectric material testers.
Novtek is a 12-year-old, $10 million company that started as a small consulting firm in test equipment applications, then became a test equipment manufacturer. Presently, Novtek specializes in flash memory and non-volatile memory production test equipment. Privately held with no venture capital and no bank debt, Novtek employs 50.
The Othello system is a fully integrated ferroelectric tester capable of 25MHz operation. The system can be used in a conventional Sawyer Tower configuration or in conjunction with a proprietary Variable Reference Sensing Amplifier (VR). According to Novtek, ferroelectric material tests, such as hysteresis, fatigue, imprint and pulse response, are all improved with the new VR technique. |
