July 11, 2012, FlexTech Alliance, San Francisco—Chan-Long Shieh from CBRITE described the efforts in moving metal oxide TFTs (MOTFT) into production on flexible plastic substrates. The issues in moving a technology from a rigid substrate to a flexible one are not insurmountable, but require much more research.

The motivation for the work is to address the changing form factors, costs, and weight for displays on mobile devices. To reduce the cost of the substrate, they are considering the development of a roll-to-roll process. The process will be additive patterning, and will require alternative deposition processing to become viable.

The existing technologies use a rigid glass substrate, which allows for easier alignment and no temperature sensitivity for the anneal steps. The problem is that the processes are expensive and planar. The free-standing sheets require self-alignment of the active devices to ensure operational functionality.

The various technologies for devices have different characteristics and use models. Silicon CMOS offers very high density and performance, and the cost per transistor is approximately zero. TFTs, on the other hand, have much higher costs per device and very low performance. The cost per area for a TFT, is much lower than the costs for CMOS. As a result, TFTs are well positioned for applications that use few transistors and lots of interconnect; displays and area sensors.

When active and passive components are mixed with high performance TFTs, the transistors exhibit a transit time that is proportional to the gate length squared. This dependence on gate length makes patterning and alignment critical. A low temperature process technology needs good mobility and stability to maintain the relationship between channel length and cost.

When MOTFTs are processed on glass, the anneal temperature may be in excess of 300C. The resulting devices are good for OLED, LCD TV, and backplanes with mobilities greater than 50 cm2/volt-second. Current on-off ratios are greater than 1010 and output resistance is over 10 MOhms at 10 uA. The device parameters show little temperature dependence and the patterned devices are stable.

The transition from glass to plastic requires a low temperature process, ideally less than 150C. One of the challenges is to produce a high-k gate dielectric at the low temperature. Material stability and low voltage operation are other areas needing ore work. The design issues for the devices include gate development, organic stop etch, and high quality interlayer dielectric materials.

One interesting material is anodized aluminum for the high-k dielectric. It is self healing, has good step coverage, low leakage, and minimum pinholes. The aluminum has low material and processing costs and seems to form a stable compound. The problem is that it needs a barrier layer to be effective and waterproof.

Patterning is another issue. The materials have to allow tradeoffs across resolution, field size, and throughput. The lithography has to image minimal line widths. In the semiconductor areas, lithography is working at the tens of nanometer range. Flat panel photolithography can achieve 3 micron resolution, screen printing is over 100 micron, and inkjet is over 20 microns.

In addition, roll-to-roll printing has to compensate for alignment and deformation as the material goes through the rollers. The process has to be self-aligned and have non-critical channel and source-drain contacts and features. Channel to gate positioning uses a positive resist, the source-drain use positive resist for liftoff, and the gate is in negative resist. The alternative is to print the features with hydrophilic silver ink.

The variables in developing the process are interface chemistry, oxygen, film thickness, deposition pressure, deposition power, and anneal time and temperature. The results show that the primary influencers are the chemistries and not the temperatures. The materials issues need additional development to achieve the final devices. Test devices show good characteristics and all device parameters are stable.