We demonstrate a light emitting p-i-n diode made of a highly aligned film of separated (99%) semiconducting carbon nanotubes, self-assembled from solution. By using a split gate technique, we create p- and n-doped regions in the nanotube film that are separated by a micron-wide gap. We inject p- and n-type charge carriers into the device channel from opposite contacts and investigate the radiative recombination using optical micro-spectroscopy. We find that the threshold-less light generation efficiency in the intrinsic carbon nanotube film segment can be enhanced by increasing the potential drop across the junction, demonstrating the LED-principle in a carbon nanotube film for the first time. The device emits infrared light that is polarized along the long axes of the carbon nanotubes that form the aligned film.
© 2010 OSA
Semiconducting single-wall carbon nanotubes (CNTs), quasi-one-dimensional materials with exciton energies that depend on their diameter, are actively explored for application in nano-optics and -photonics . Several studies have addressed the fundamentals of charge carrier and phonon dynamics, exciton formation as well as light absorption and emission in optoelectronic devices made of a single CNT [2–12]. Just recently, it has been demonstrated that it is possible to create such a fundamental photonic building block as the light emitting p-i-n junction with a single CNT .
The light emission properties of CNT films and networks [14–16] have attracted current interest beyond single CNT applications as quantum light sources . CNT films can be readily assembled from solution and could enable scaling up the amount of light that can be emitted or detected by a CNT device. Moreover, CNT films achieve consistent output by averaging-out the heterogeneities among individual CNTs and also make optoelectronic devices more robust against failure. Ultimately, self-assembled and highly aligned arrays with many CNTs in parallel are expected to preserve the polarization provided by the one-dimensional character of individual CNTs. Since emission and absorption of light that is polarized parallel to the long axis of a CNT occurs with higher efficiency, an aligned CNT film could be a promising building block for future laser applications.
In this paper, we report the realization of a light emitting p-i-n diode from a highly aligned film of semiconducting carbon nanotubes that operates in the near-infrared spectral range. A split gate design allows for tuning both the rectifying electrical behavior of the diode and its light generation efficiency. The CNT film diode emits light that is polarized along the device channel, a direct consequence of the high degree of CNT alignment in the film.
2. Materials, device fabrication and experimental methods
We deposited CNT thin films on a Si(p++)/SiO2(100nm) substrate via a novel evaporation-driven, self-assembly technique  from a suspension of 99% semiconducting (arc discharge) carbon nanotubes with diameters ranging from 1.3 to 1.7 nm, separated by density-gradient ultracentrifugation . With this deposition technique, CNTs align in parallel at the contact line between the solution and a vertically immersed planar substrate. The slip-stick motion of the contact line during evaporation produces periodic arrays of regular CNT thin film stripes that cover the entire substrate (i.e. an area of around 50 mm2 in the present case), each having a width of about 10 μm and a height of 1 to 6 nm, the latter depending on the degree of CNT bundling (see Fig. 1(b) ). The CNT stripe formation relies solely on self organization without the need for etching or lithography technique to obtain the patterns. These CNT films have been extensively characterized in situ by optical and electrical techniques as well as by SEM and AFM . As it has been demonstrated in reference , it is possible to fabricate devices that incorporate a single or even multiple parallel CNT stripes.
On top of the CNT film contacted by source and drain electrodes (Ti = 1 nm/Pd = 40 nm/Au = 20 nm), we deposited 30 to 50 nm of Al2O3 by means of atomic layer deposition followed by Ti top gates (thickness: 15 to 35 nm), defined by electron beam lithography. As sketched in Fig. 1(a), the top gates enable electrostatic doping of the CNT film segments underneath. Applying different voltages to the gates creates a local potential drop across the film that is spatially confined to the extent of the gate separation (~1 μm, see Fig. 1(b)). The channel length L C between source and drain contacts ranges from 4 μm to 10 μm and exceeds the average CNT length (~1 μm) in order to avoid shorting the channel with residual metallic nanotubes. Hence, electronic transport along the channel relies on percolation. Figure 1(b) shows an SEM image of a CNT film diode with L C= 4 μm.
Electrical transport and optical measurements were performed in a vacuum chamber at a pressure level of around 5 x 10−6 Torr (since CNT channels are protected by Al2O3 thin films we expect that devices would operate similarly under ambient conditions). Electroluminescence emission of the devices passed the sapphire window of the vacuum chamber, a 20x or 50x microscope objective lens, a transmission grating, a field lens, the detector window, and a short-pass filter (cut-off wavelength of 2150 nm or 2000 nm) before being recorded by a liquid-nitrogen cooled HgCdTe sensor array.
3. Results and discussion
The electrical transfer characteristics of two devices with different channel lengths (4 μm and 6 μm) are shown in Fig. 2(a) and (b) . The almost symmetric, ambipolar transfer characteristics of the devices indicate that both electrons and holes can be injected into the channel with about the same efficiency. On-state sheet resistances are of the order of 1 MΩ/□ while on/off current ratios are as high as 100. As compared to the single nanotube p-i-n diodes , the CNT film LEDs sustain much higher drive currents and are more persistent against device failure induced by electrostatic discharge. The hysteresis is likely to be caused by trapped charges in the gate oxides that may affect the optical emission of the CNT film to the extent that they change the actual potential landscape in the device.
In Fig. 2(c) and (d), we show the corresponding electrical output characteristics of the two devices. By stepping up voltages of opposite polarity to the two split gates, we observe that the reverse current decreases and the forward current increases at a fixed source-drain voltage, as indicated by the arrows in Fig. 2(d). Hence, by increasing the potential drop between the split gates, we obtain a gradual increase of the rectifying behavior in the CNT film diode, as expected. As L C increases from 4 to 6 μm, we observe that the split gate effect becomes more pronounced and the attainable on-current decreases. This is analogous to the observation of an increasing on/off current ratio and a decreasing on-current with increasing L C in devices made out of the same film but equipped with a single global top gate .
For convenience, the terms “forward bias” and “reverse bias” refer to standard p-n diodes. However, the potential landscape of conduction and valence levels in these devices is far more complex than in a standard p-n diode, and we do not expect the rather ideal behavior that is observed in p-i-n diodes made of single CNTs [13, 19]. In the present case, CNTs with different diameters and chiralities are contacted, with many of them bundled, and charge carriers have to overcome various tube-to-tube contacts in order to cross the channel. Also, the height of the Schottky barriers between the metallic source/drain contacts and the CNTs are highly variable as external electric fields are applied . As a result, we are not able to deliver a quantitative picture of the nanometer-scale potential landscape in CNT percolation film devices. Further research is needed to obtain a better understanding of the complex band structure in these systems. Nevertheless, in the case the voltage range is restricted to moderate values of V DS between −5 V and +5 V, the CNT film diodes can be operated in a purely rectifying mode as shown exemplarily in Fig. 2(e) for a CNT film diode with L C= 10 μm.
Next, we discuss the electroluminescence properties of the CNT thin film devices. In general, the radiative decay of excitons is responsible for light emission in CNTs. Excitons can be electrically generated in CNTs via (I) simultaneous injection of electrons and holes that form excitons and radiatively recombine (giving rise to threshold-less ambipolar electroluminescence) or (II) accelerating electrons or holes to energies sufficient to create excitons, i.e., the so-called impact excitation (giving rise to unipolar electroluminescence with threshold characteristics) . By using a p-i-n design, we aim to improve the control of simultaneous injection of electrons and holes in order to increase the efficiency of the threshold-less ambipolar electroluminescence, making the CNT film a more efficient light emitter.
We now turn to the measured light emission characteristics of the CNT film diodes. The devices emit a broad electroluminescence band in the near infrared with a maximum spectral intensity at about 0.6 eV , which is in agreement with the expected value based on the known CNT diameter distribution of the sample. Figure 3(a) shows the electroluminescence intensity of a device integrated over wavelength as a function of source-drain voltage. As a reference, we plot in the inset of Fig. 3(a) the electrical transport characteristics of the same device. For a fixed V DS, as we increase the potential drop across the CNT film stepwise from 0V/0V to + 5V/-5V, we observe a substantial increase of light intensity of up to a factor of 16 in the forward direction. In contrast, despite the high currents achieved (see inset Fig. 3(a)), we only observe relatively weak electroluminescence emission if we sweep V DS in the reverse direction. Also, we do not observe a significant influence of the split gate in the intensity characteristics of the emitted light if the device is biased in the reverse direction.
In order to investigate whether the light generation efficiency is improved by creating a potential drop across the CNT film, we plot in Fig. 3(b) the integrated electroluminescence intensity as a function of the injected electrical power, P EL=I DS·V DS for several voltage configurations. With the device biased in the reverse direction, we obtain linear power dependences with almost identical slopes, independent of the split gate voltages applied. If the device is biased in the forward direction, we observe that the slopes of the intensity curves are steeper than the curves in the reverse mode. This indicates that the device generates light more efficiently if biased in the forward direction, even without split gate voltages applied. This effect is associated with built-in non-uniformities of the CNT film and the potential landscape of individual devices. We found that a certain level of asymmetry also exists in the electrical transport behavior (see for example the transport at 0V/0V in Fig. 2(e)), making devices operate more efficiently for a certain choice of (source/drain) bias conditions.
More importantly, if the split gate voltages are tuned from 0V/0V to +5V/-5V and the device is swept in the forward direction, we observe a strong increase of light generation efficiency (Fig. 3(b)). The power dependence of the most efficient configuration (+5V/-5V) can be captured adequately by fitting a quadratic form to the measured data points. The somewhat non-linear power dependence observed might be due to either higher order electrostatic field contributions (across the junction) or higher order excitonic effects (in the CNTs); the investigation of both effects is beyond the scope of this paper. At P EL = 150 μW, the light generation efficiency has increased by a factor of 4 as compared to the reverse mode.
We attribute the observation of the improved light generation efficiency in the forward mode to an enhancement of ambipolar recombination (rather than impact excitation) in the intrinsic CNT film segment. The threshold characteristics of the device are consistent with our interpretation. While the CNT film diode generates light virtually threshold-less in the forward mode (see inset Fig. 3(b)), it displays the threshold behavior associated with impact excitation when biased in the reverse mode. By comparing to single CNT diodes attaining conversion efficiencies of the order of 10−4 photons per electron-hole pair (exciton) , we estimate that the conversion efficiency of the CNT film diode is lowered by at least one order of magnitude. This result can be understood by considering that: (I) the CNT film diode is a percolation device relying on non-ideal nanotube-nanotube contacts; (II) the CNT material has a much higher defect density because it has been solution-processed and (III) the presence of remaining metallic nanotubes in the film effectively quench light emission due to non-radiative energy transfer. Nevertheless, by comparing the light output of a CNT film diode and a diode made of a single CNT as function of drain bias (see Fig. 4(a) ), we obtain an increase of light output by factors of up to 15, demonstrating the scaling potential of CNT thin film LEDs.
Finally, we investigate the polarization dependence of the electroluminescence emission of the CNT film diode. Based on SEM (see Fig. 1(b)) and Raman imaging, individual CNTs (bundles) forming the film have been measured to lie within a range of 5° with respect to one another . Since single CNTs are known to emit light that is strongly polarized in the direction of their long axes, we expect that the high degree of CNT alignment in the film translates into a pronounced light polarization effect, as indicated schematically in Fig. 1(a). Indeed, Fig. 4(b) clearly shows that the electroluminescence is linearly polarized and intensity maxima are obtained if the transmission direction of the polarizer is aligned with the CNT direction. The cosine-squared fits reveal intensity ratios I max/I min (i.e. ratios with polarizer at 0°/180° over 90°/270° in Fig. 4(b)) of around 4, regardless of whether or not the device is biased by the split gates. The degree of polarization is of the same order as the values obtained in electroluminescence [2, 7], photoluminescence and Raman  as well as photoconductivity measurements [3, 4] performed on single nanotubes, demonstrating the extraordinary level of alignment in the CNT film. We should point out that the light polarization is preserved even if the CNT film diode is operated in the reverse mode. This is not unexpected, since the polarization dependence of the electroluminescence measurement reflects the nature of the exciton recombination rather than the exciton creation. Since we do not observe light emission from the CNT-metal-contacts that are covered by the metallic gates, the polarization dependence in both forward and reverse mode reflects solely the alignment of the CNTs in the “active area” between the split gates (see Fig. 1).
4. Summary and conclusions
We have realized polarized light-emitting diodes from highly aligned, separated semiconducting carbon nanotube films that show tunable light generation efficiencies and threshold-less light emission characteristics. Future challenges lie in further enrichment of the solution-processed CNT material for device fabrication, enabling shorter device channels and, ultimately, allow for building devices that do not rely on percolation. Also, avoidance of excess chemical treatment and processing will decrease the defect concentration of the CNT material and help to further improve the light generation efficiency.
We thank Damon Farmer (IBM Research, Yorktown Heights, NY) for preparing gate dielectric thin films by atomic layer deposition. We acknowledge expert technical support by Bruce A. Ek (IBM Research, Yorktown Heights, NY). M. K. acknowledges support from NSF (DMR # 0705131). A. A. G. and M. C. H. acknowledge support from the National Science Foundation (DMR-0520513, EEC-0647560, and DMR-1006391) and the Nanoelectronics Research Initiative.
References and links
1. P. Avouris, M. Freitag, and V. Perebeinos, “Carbon-nanotube photonics and optoelectronics,” Nat. Photonics 2(6), 341–350 (2008). [CrossRef]
2. J. A. Misewich, R. Martel, P. Avouris, J. C. Tsang, S. Heinze, and J. Tersoff, “Electrically induced optical emission from a carbon nanotube FET,” Science 300(5620), 783–786 (2003). [CrossRef] [PubMed]
3. M. Freitag, Y. Martin, J. A. Misewich, R. Martel, and P. Avouris, “Photoconductivity of Single Carbon Nanotubes,” Nano Lett. 3(8), 1067–1071 (2003). [CrossRef]
4. K. Balasubramanian, Y. Fan, M. Burghard, K. Kern, M. Friedrich, U. Wannek, and A. Mews, “Photoelectronic transport imaging of individual semiconducting carbon nanotubes,” Appl. Phys. Lett. 84(13), 2400–2402 (2004). [CrossRef]
5. M. Freitag, V. Perebeinos, J. Chen, A. Stein, J. C. Tsang, J. A. Misewich, R. Martel, and P. Avouris, “Hot Carrier Electroluminescence from a Single Carbon Nanotube,” Nano Lett. 4(6), 1063–1066 (2004). [CrossRef]
6. J. Chen, V. Perebeinos, M. Freitag, J. Tsang, Q. Fu, J. Liu, and P. Avouris, “Bright infrared emission from electrically induced excitons in carbon nanotubes,” Science 310(5751), 1171–1174 (2005). [CrossRef] [PubMed]
7. D. Mann, Y. K. Kato, A. Kinkhabwala, E. Pop, J. Cao, X. Wang, L. Zhang, Q. Wang, J. Guo, and H. Dai, “Electrically driven thermal light emission from individual single-walled carbon nanotubes,” Nat. Nanotechnol. 2(1), 33–38 (2007). [CrossRef]
8. L. Xie, H. Farhat, H. Son, J. Zhang, M. S. Dresselhaus, J. Kong, and Z. Liu, “Electroluminescence from suspended and on-substrate metallic single-walled carbon nanotubes,” Nano Lett. 9(5), 1747–1751 (2009). [CrossRef] [PubMed]
9. M. Steiner, M. Freitag, V. Perebeinos, A. Naumov, J. P. Small, A. A. Bol, and P. Avouris, “Gate-variable light absorption and emission in a semiconducting carbon nanotube,” Nano Lett. 9(10), 3477–3481 (2009). [CrossRef] [PubMed]
10. M. Steiner, M. Freitag, V. Perebeinos, J. C. Tsang, J. P. Small, M. Kinoshita, D. Yuan, J. Liu, and P. Avouris, “Phonon populations and electrical power dissipation in carbon nanotube transistors,” Nat. Nanotechnol. 4(5), 320–324 (2009). [CrossRef] [PubMed]
11. M. Freitag, M. Steiner, A. Naumov, J. P. Small, A. A. Bol, V. Perebeinos, and P. Avouris, “Carbon nanotube photo- and electroluminescence in longitudinal electric fields,” ACS Nano 3(11), 3744–3748 (2009). [CrossRef] [PubMed]
12. N. M. Gabor, Z. Zhong, K. Bosnick, J. Park, and P. L. McEuen, “Extremely efficient multiple electron-hole pair generation in carbon nanotube photodiodes,” Science 325(5946), 1367–1371 (2009). [CrossRef] [PubMed]
13. T. Mueller, M. Kinoshita, M. Steiner, V. Perebeinos, A. A. Bol, D. B. Farmer, and P. Avouris, “Efficient narrow-band light emission from a single carbon nanotube p-n diode,” Nat. Nanotechnol. 5(1), 27–31 (2010). [CrossRef]
14. M. Engel, J. P. Small, M. Steiner, M. Freitag, A. A. Green, M. C. Hersam, and P. Avouris, “Thin film nanotube transistors based on self-assembled, aligned, semiconducting carbon nanotube arrays,” ACS Nano 2(12), 2445–2452 (2008). [CrossRef]
15. E. Adam, C. M. Aguirre, L. Marty, B. C. St-Antoine, F. Meunier, P. Desjardins, D. Ménard, and R. Martel, “Electroluminescence from single-wall carbon nanotube network transistors,” Nano Lett. 8(8), 2351–2355 (2008). [CrossRef] [PubMed]
16. J. Zaumseil, X. Ho, J. R. Guest, G. P. Wiederrecht, and J. A. Rogers, “Electroluminescence from electrolyte-gated carbon nanotube field-effect transistors,” ACS Nano 3(8), 2225–2234 (2009). [CrossRef] [PubMed]
18. M. S. Arnold, A. A. Green, J. F. Hulvat, S. I. Stupp, and M. C. Hersam, “Sorting carbon nanotubes by electronic structure using density differentiation,” Nat. Nanotechnol. 1(1), 60–65 (2006). [CrossRef]
19. J. U. Lee, P. P. Gipp, and C. M. Heller, “Carbon nanotube p-n junction diodes,” Appl. Phys. Lett. 85(1), 145–147 (2004). [CrossRef]