Manufacturing Bits: June 13

FeFET market heats up; better ferroelectrics; 2D ReRAMs.


FeFET biz heats up
The ferroelectric FET (FeFET) market is heating up.

One company, Ferroelectric Memory Co. (FMC), has been developing FeFETs, a new memory type for use in standalone and embedded applications. Now, Imec is also developing FeFETs in both planar and vertical varieties.

Imec’s FeFET (Source: Imec)

For years, several chipmakers have been shipping traditional ferroelectric memories or FRAMs, mostly for embedded applications. FRAM is based on a 1T-1C memory cell. The ferroelectric film is implemented in the capacitor, but the technology is difficult to scale. In contrast, FeFETs are based on the ferroelectric properties of hafnium oxide. The ferroelectric replaces the gate dielectric of a CMOS transistor.

Meanwhile, at the recent 2017 Symposia on VLSI Technology and Circuits, Imec presented a paper about the development of a vertically-stacked FeFET. The ferroelectric device is an aluminum-doped, hafnium-oxide technology for NAND applications. The technology has various characteristics, such as low power consumption, fast switching speeds, scalability and high retention rates.

Ferroelectric materials consist of crystals. They have spontaneous polarization. In other words, ferroelectric technology can be in one of two states, which can be reversed with a suitable electric field, according to Imec. The breakdown of the interfacial layer and low retention characteristics presented challenges for FRAMs. But the ferroelectric phase in hafnium oxide has created a buzz in the industry.

“With HfO2, there is now a material with which we can process ferroelectric memories that are fully CMOS compatible. This allows us to make a ferroelectric FET (FeFET) in both planar and vertical varieties,” said Jan Van Houdt, Imec’s chief scientist for memory technology, on Imec’s Web site.

“We are working to overcome some of the remaining issues, such as retention, precise doping techniques and interface properties, in order to stabilize the ferroelectric phase,” Van Houdt said. “We are now confident that our FeFET concept has all the required characteristics. It is, in fact, suitable for both stand-alone and embedded memories at various points in the memory hierarchy, going all the way from non-volatile DRAM to flash-like memories. It has particularly interesting characteristics for future storage-class memory, which will help overcome the current bottleneck caused by the differences in speed between fast processors and slower mass memory.”

Better ferroelectrics
The Department of Energy’s Lawrence Berkeley National Laboratory (Berkeley Lab) has expanded the range of functional temperatures for traditional ferroelectrics materials.

To achieve this feat, researchers have created the first-ever polarization gradient in a thin film. For this, researchers created a strain and chemical gradient in a 150nm film of barium strontium titanate. The polarization varied from 0 to 35 microcoulombs per centimeter squared across the material, according to researchers.

To develop the strain, they used an epitaxial tool. With the tool, a crystalline film is grown on a substrate with a mismatch in the lattice structure. “Traditional physics and engineering textbooks wouldn’t have predicted this observation,” said Lane Martin, faculty scientist at Berkeley Lab’s Materials Sciences Division and UC Berkeley associate professor of materials and engineering, on the university’s Web site.

“Creating gradients in materials costs a lot of energy—Mother Nature doesn’t like them—and the material works to level out such imbalances in whatever way possible. In order for a large gradient like the one we have here to occur, we needed something else in the material to compensate for this unfavorable structure. In this case, the key is the material’s naturally occurring defects, such as charges and vacancies of atoms, that accommodate the imbalance and stabilize the gradient in polarization,” Martin said.

“The new polarization profile we have created gives rise to a nearly temperature-insensitive dielectric response, which is not common in ferroelectric materials,” Martin said. “By making a gradient in the polarization, the ferroelectric simultaneously operates like a range or continuum of materials, giving us high-performance results across a 500-degree Celsius window. In comparison, standard, off-the-shelf materials today would give the same responses across a much smaller 50-degree Celsius window.”

Lanza Laboratory, the Massachusetts Institute of Technology, Stanford University and Harvard University have developed a resistive random access memories (ReRAM) using 2D material.

The 2D material involves a compound based on graphene/hexagonal-boron-nitride/graphene (G/h-BN/G). The h-BN material serves as the resistive switching medium. The 2D-based ReRAM device was fabricated using traditional tools. Researchers also devised a model for the multi-filamentary conduction in h-BN. In the model, the device shows reproducible resistive switching.