The University of Florida has developed an organic thin-film transistor (TFT) for active matrix OLED displays.
Because OLED pixels are current driven, they require TFTs with a far lower on-resistance than the TFTs used in LCD backplanes.
High mobility, a narrow channel and high gate capacitance promote low on-resistance and low gate voltage operation.
With only the poor mobility of an organic semiconductor available - even the best are considerably worse than even polysilicon - the Florida researchers have concentrated on increasing gate area and cutting dielectric thickness to increase capacitance.
What they have developed is a novel stacked vertical structure (see diagram below).
The layered approach automatically gives a large gate area, and a thin channel can be achieved by spin-coating, or other techniques, without resorting to fine-geometry lithography.
200µm resolution is sufficient to lay out the drain electrode which is the most demanding feature, compared with 25µm needed for some lateral TFTs.
"A conventional TFT would need high-resolution imaging to get a narrow channel for high current and low voltage operation," Professor Andrew Rinzler of the University told Electronics Weekly.
Like many TFTs, the Florida transistor has a bottom gate structure which means it has to affect the channel through the source .
"You need a discontinuous source electrode," said Rinzler. "If it was continuous, especially if it was metal, you would screen away the gate field."
To allow the gate to work, the source is constructed from a random open mesh of carbon nanotubes, forming an electrode that is electrically semi-transparent.
"We are using dilute intimately-connected nanotubes as the source electrode to inject carriers into the channel," said Rinzler. "The reason that it works is that the nanotubes form a Schottky barrier with the polymer channel material. The gate field modulates the height and width of the Schottky barrier, and that modulation allows the carriers to be injected into the layer or not."
The gate is aluminium on a glass substrate, with its upper surface oxidised to make a thin (5nm) Al2O3 gate insulator.
On its own, this would deliver 1.7µF/cm2.
However, for added reliability, a 5nm layer of the insulator benzocyclobutene (BCB) is spin-coated over this, reducing gate capacitance to 350nF/cm2.
According to the University, there is a considerable scope to improve the gate stack.
The channel material is a 480nm thick layer of dinaphtho-[2,3-b:20,30-f ]thieno[3,2-b]thiophene (DNTT), a recent invention first synthesised by Dr Tatsuya Yamamoto of chemical firm Nippon Kayaku and Professor Kazuo Takimiya of Hiroshima University
DNTT is a small molecule semiconductor with high hole mobility (2.9cm2/Vs) and more stability in air than well-known pentacene.
Test devices made with pentacene lost an order of magnitude of performance after three months, compared with DNTT versions that lost half after eight months.
Its other advantage is that its work function is negative enough to form the necessary barrier to hole injection at its interface with the nanotubes.
The result is a device that operates at low voltage.
Swinging the gate from -2V to +2V delivers a 105 on/off ratio with the drain anywhere between -1V and +1V.
If an on/off ratio of 104 can be tolerated, an on-current density of 110mA/cm2 can be achieved with -3V on the drain.
"On-current is sufficient at low-enough voltages to drive an OLED - which is not true of lateral polymer-based TFTs. Around 1mA is needed for a blue pixel," explained Rinzler. "Off-current suffers a little bit, but difference is negligible, and the power lost in leakage is more than compensated for by lower losses in the TFT when it is on."
With the lower drain voltage, the on/off ratio stays at 105 up to 50mA/cm2, and above 104 to over 110mA/cm2.
The team claims its device beats other developments by 3.9x in terms of current density, and predicts that, driving a pixel with 4cd/A output at 25mA/cm2, the OLED would deliver 1000cd/m2, compared with around 300cd/m2 maximum form a standard TV.
20 devices were fabricated, and showed only small variations in current density and threshold voltage, and temperature cycling shifted gate threshold voltage only a few percent.
Are there drawbacks with this device?
There are.
However, Rinzler is obliged not to discuss them as a further development has dealt with them, and publication of these results is imminent.
"There are challenges that we have overcome, but I cannot say anything because a paper due out very shortly," he said.
Further scope for development, according to the team, could be offered by the optical transparency of the nanotube source electrode.
This absorbs only 2% of light and could form part of a light-emitting transistor.
The original work was reported in the American Chemical Society journal Nano Letters.
Nanotubes as channels in transistors
Carbon nanotubes are more usually thought of as a channel material in nano-scale transistors.
For example, several years ago IBM built a CMOS ring oscillator on a single semiconducting single-walled carbon nanotube.
In this case, IBM made the p and n-mosfets along the nanotube by imparting n and p-channel characteristics to the tube locally by using aluminium and palladium gates respectively, all over an aluminium oxide gate insulator.
The five stage oscillator ran at 53MHz despite almost no optimisation, and THz operation was extrapolated.
However, although carbon nanotubes are easily grown - fine for bulk uses like the University of Florida TFT project - they are not easily grown in exactly the right place for nano transistors and generally have to be manipulated into position.
Interest in nm-scale nanotube transistors was diverted somewhat by the discovery of graphene by the University of Manchester.
Graphene is a flat carbon lattice with characteristics similar to nanotubes, but it can be patterned with lithography.
True, the best graphene still has to be flaked from graphite and manipulated into position, but grown-in-situ graphene is improving.
The photo at the top shows the transistor on a glass substrate. The black spots are the gold drain array, gold circles at the side are gate contacts (sitting on the aluminium gate), and the circles at the bottom contact the source. The DNTT channel semiconductor is yellow.