Optoelectronics with Carbon Nanotubes

3. Electroluminescence mechanism and characteristics
The left image in Figure IV-2 (a) depicts the infrared emission from a device when the
gate electrodes are biased at VGS1 = -8 V, VGS2 = +8 V and a constant current of IDS = 240 nA is
driven through the SWNT. As discussed in the previous chapter, electrically excited light
emission from semiconducting CNTs can be produced under ambipolar 73, 76 or unipolar
operation. In the former case, both electrons and holes are injected simultaneously into the tube
and their radiative recombination generates light. In the second case, a single type of carriers,
i.e. either electrons or holes, accumulate kinetic energy in a high-field region within the device to
generate excitons by means of impact excitation 71, 84, 85 . The fact that no light is emitted when
our devices are operated under unipolar conditions (VGS1 = VGS2 = -8 V; hole current) – right
image in Fig. IV-2 (a) – shows that they are ambipolar light-emitters. This is the behavior we
would generally expect an LED to exhibit. It is essential when fabricating devices that the back-
gate sweep show a good ambipolar behavior (not shown), i.e., both types of carriers must be
injected into the channel efficiently, in order for the devices to show this type of emission. Since
the recombination rate in the ambipolar case is limited by minority carriers, this gives additional
evidence for the ambipolar nature of our observation.
Figure IV-2 (b) shows that the emission intensity is proportional to the current (i.e. carrier
injection rate), which also confirms that the device is operating as an ambipolar emitter. Recall
that the EL intensity from impact excitation in unipolar emission grows as the second to third
power of the current, as the increased field makes the exciton generation more efficient (Figure
III-12). Moreover, the emission threshold with respect to the current is zero, and the signal is
still detectable at IDS as low as ~10 nA in many of our devices. This is in contrast to all previous
EL studies 71, 76, 83, 111 , where typically two orders of magnitude higher current levels are required
to obtain light emission of comparable intensity. Moreover, the voltage drop across the intrinsic
region is in the order of the bandgap (~1 V; see Figure IV-1 (c)), and therefore also 5–10 times
smaller than in other devices, resulting in an up to 1000 times smaller power dissipation overall.
Under typical operation conditions, we estimate a power density of only ~0.1 W/m in the tube,
compared to 10–100 W/m in other devices. It is hence clear that the CNT diodes are operated in
an entirely different regime from all other electrically-driven CNT light-emitters to date. In fact,
the power density is comparable to what is typically used in photoluminescence (PL)
70
71, 85