Noise spectroscopy of transport properties in carbon nanotube field-effect transistors

Transcription

1 CARBON 53 (2013) Available at journal homepage: Noise spectroscopy of transport properties in carbon nanotube field-effect transistors V.A. Sydoruk a, M.V. Petrychuk b,a.ural c, G. Bosman c, A. Offenhäusser d, S.A. Vitusevich a, *,1 a Peter Grünberg Institute, Forschungszentrum Jülich, Germany b Taras Shevchenko National University, Kiev, Ukraine c Department of Electrical and Computer Engineering, University of Florida, USA d Peter Grünberg Institute/Institute of Complex Systems (PGI-8/ICS-8): Bioelectronics, Forschungszentrum Jülich and JARA Fundamentals of Future Information Technology, Jülich, Germany ARTICLE INFO ABSTRACT Article history: Received 3 July 2012 Accepted 27 October 2012 Available online 2 November 2012 Transport properties of single-walled carbon nanotube (CNT) structures with Pd contacts were studied using noise spectroscopy. The high values of the mobility and low noise level are characteristic of high-quality CNT material. The detailed analysis of the transport and noise properties of the CNT structure with back gate topography allows us to study the transport determined by Schottky barriers and by pure CNT channel conductivity and to establish their separate contribution to the total conductivity of the structure. It was demonstrated that at small gate overdrive the main source of flicker noise is related to the Schottky barriers of the CNT FETs. With increasing gate voltage, the magnitude of flicker noise decreases and at a certain gate voltage it is only determined by the transport properties of carbon nanotubes with a noise level lower by one order of magnitude. In contrast to previous studies where flicker noise determined the excess noise of CNT-based structures, we registered generation recombination noise components in our structures and studied their behavior in a wide temperature range. This allowed us to investigate the origin of traps capturing the carriers, which considerably affects the noise and transport properties of CNT structures. Ó 2012 Elsevier Ltd. All rights reserved. 1. Introduction Carbon nanotubes (CNTs) demonstrate unique properties, therefore they represent an interesting class of materials for research and applications. CNTs are materials with a high degree of ordering of carbon atoms demonstrating higher carrier mobilities compared with those of conventional semiconductor materials. CNTs with semiconductor properties have become objects of increasing interest as materials for the development of one of the most important devices in nanoelectronics the field-effect transistor (FET) with advanced characteristics [1]. Current technology for the fabrication of CNT-based structures does not allow high-quality contacts to CNTs, therefore transport properties are mainly determined by contact phenomena. In conventional semiconductor devices, problems of contacting are usually avoided by replacing metal contacts with heavily doped regions of semiconductors, which is difficult in the case of CNTs [2]. Therefore, direct metal semiconductor contacts in devices fabricated on the basis of CNTs have relatively wide Schottky barriers. * Corresponding author. address: s.vitusevich@fz-juelich.de (S.A. Vitusevich). 1 On leave from Institute of Semiconductor Physics, NASU, Kiev, Ukraine /$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.

2 CARBON 53 (2013) The major role of the Schottky contacts in CNT FETs was previously suggested on the basis of results from studying the ambipolar electrical transport [3], the conductivity of logic circuits with CNT FETs [4], investigating CNT FETs using a scanning gate microscope [5], etc. Later, Heinze et al. [6] demonstrated that CNT FETs operate as Schottky barrier transistors. The authors suggested that transistor properties are determined by varying the contact resistance rather than the channel conductance. Additionally, electrical contacts to metallic CNTs are also found to be diameter-dependent especially for nanotubes with small diameters. At the same time, the investigation of CNT FETs with Pd, Ti, Al contacts and with different diameters of nanotubes showed that Pd is the best metal to improve p-cnt FET performance due to its high work function [7]. By fitting the experimentally obtained data [8] using the distributed contact model, Solomon showed that it was possible to achieve improved performance of the contacts to the nanotube by establishing a long ballistic mean free path under the contacts [9]. However, ab initio electronic structure studies (with large-scale transport calculations) revealed that, in addition to the type of contact metal, the interface morphology and the length of the contact region play an important role as well [10]. The investigation of CNT FETs with Pd and Rh contact metallization clearly showed the dependence of the electrical properties of semiconducting CNTs on their diameters. The dependence was explained by the formation of Schottky barriers in the case of nanotubes with diameters below 1.6 nm [11]. To realize the full potential of CNT material with unique properties, the transport characteristics and influence of contact phenomena on the conductivity of whole system in different functional regimes have to be studied. Noise spectroscopy is a powerful method for studying the performance of materials, structures and transport phenomena in complex systems, where contributions of different components of the system to the system properties have to be studied separately. The efficiency of noise spectroscopy increases at the nanoscale because fluctuation phenomena contain important information about the materials, which may not be accessible by conventional methods. Usually, the carbon nanotubes exhibit 1/f-noise in a wide frequency range, which gives an opportunity to understand phenomena related to 1/f-noise in 1D conducting channels. The first results on CNTs 1/f-noise were reported by Collins et al. [12]. An anomalously large bias-dependent 1/f noise was observed. The crucial role of the contacts to CNTs was shown in the work of Kim et al. [13], where the low-frequency noise of individual multi-walled CNTs was investigated with Ti/Au, Cr/Au, Pt/Au, and Pd/Au electrodes. Pd- and Pt-contacted devices showed a much smaller noise level than Ti- and Cr-contacted devices. The noise level is an important parameter, which has to be minimized in most cases. The noise level of different materials and systems can be compared by estimating the dimensionless Hooge parameter [14]. In the paper by Collins et al. [12], an estimation of Hooge s constant, a H, was made taking the number of carriers as the number of atoms. The obtained value of a H = 0.2 was about 100 times higher than the typical value proposed by Hooge for different materials [14]. Ishigami et al. [15] found that Hooge s empirical rule adequately describes the low-frequency noise in CNT FETs with a H = (9.3 ± 0.4) Lin et al. [16] showed that Hooge s constant in nanotubes is comparable to most bulk materials and that the noise amplitude can be quite significant because of the small number of carriers. By investigating the temperature dependence of the resistivity of CNT films, two mechanisms (hopping and fluctuation-induced tunneling) of current formation were demonstrated by Behnam et al. [17] The authors analyzed the temperature dependence of the noise amplitude and extracted the density of fluctuators that are responsible for the 1/f-noise as a function of their energy using a method demonstrated in the work of Tobias et al. [18]. Investigation of transport properties in bundles of carbon nanotubes with metallic conductivity was performed by Danilchenko et al. [19] in a wide range of temperatures and revealed three mechanisms of transport: hopping conductivity, Luttinger liquid conductivity, and diffusion conductivity. Despite the progress in CNT technology, mechanisms of transport in the structures fabricated on the basis of CNTs with a semiconductor type of conductivity are still under discussion. One of the approaches to study performance as well as the defect structure of materials and systems and their role in transport properties is the measurement of generation recombination (GR) noise components in a wide temperature range. However, up to now GR noise components in CNT FETs have not been observed mainly due to a high level of excess flicker noise. In the present paper, we report results demonstrating GR noise component registration in noise spectra of CNT-based structures. The data enable information to be obtained about the influence of traps on noise and transport properties in semiconductor CNTs. Studies of single-walled CNT FETs with back gate topography, which allows control over the transport regimes in the devices using different gate voltages, make it possible to separate transport determined by the Schottky contacts and that determined by the CNT material itself. Using Pd (Cr is used as an adhesion layer underneath Pd) as a contact metal, which has a work function of ev, allowed us to investigate transport in FET samples using noise spectroscopy at different gate voltages and find a regime with negligible influence from the Schottky contacts. In this regime, transport properties are determined by the CNT material. 2. CNT structure design The structures under study were fabricated on the basis of individual single-walled CNTs with diameters approximately 1 nm grown on a Si/SiO 2 substrate by using the chemical vapor deposition (CVD) method. After the growing of CNTs, the network of paired contacts (with various distances between them) was deposited. Thus, the channel of FETs consists of the individual nanotubes contacted by two Cr/Pd electrodes which operate as source and drain contacts. The width of the electrodes is 100 lm and the distance between them 6 lm. Typical CNT FET structures are shown in Fig. 1 with a total length of the CNT channel of (6 ± 1) lm. The heavily doped Si substrate was used as the back gate of the FETs. 3. Experimental details A low-noise home-made noise measurement setup was used. A lead-acid battery was used to apply voltage to the sample,

3 254 CARBON 53 (2013) Results and discussion In order to study the transport mechanisms in the FETs, the current voltage and noise characteristics of CNT FET samples were measured at different temperatures and applied drain and gate voltages. The maximum value of the transconductance, g m, is found to be as high as 0.1 ls, which is typical taking into account the thick dielectric layer with thickness t = 500 nm between the gate and the CNT channel. The effective mobility, l, of the CNT can be estimated using the following equation for long-channel FETs at low drain source voltage [20]: l ¼ L2 DS g m C G V DS ; ð2þ Fig. 1 Scanning electron microscope image of a typical CNT FET. which can be controlled using a variable resistor of 1 kx. Such a power supply system allowed us to reduce noise pickups during the measurements. The samples were connected to the power source in series with load resistance, R load. A high-precision, low-inductance resistance box, produced by IET Labs Inc. (formerly manufactured by GenRad), was used to select the load resistance. The resistance can be changed in the range from 1 X to 1 MX. We measured the voltage on the sample, V S, and the total voltage, V M, using a multi-meter before as well as after the noise measurements. This allows us to monitor the stability of the sample and calculate the current through the sample. The current through the sample, I S, was estimated by following formula: I S ¼ V M V S : ð1þ R load For CNT FETs measurements, we used R load =5kX. The noise signal from the sample was amplified by a home-made preamplifier with low noise level corresponding to the thermal noise of 140 X resistance at frequency 100 Hz and a gain of 177 as well as by a commercial amplifier (ITHACO 1201) with variable gain. The input impedance of the preamplifier in our measurement system can be estimated to be 1 MX in the AC regime (above 0.3 Hz). The intrinsic input-referred thermal noise of preamplifier and ITHACO amplifier is measured as V 2 Hz 1 and V 2 Hz 1, respectively. The noise spectra were registered using a dynamic signal analyzer (HP35670A) and the data were transferred through the GPIB interface to the computer for further analysis. The noise characteristics were measured in a wide frequency range, from 1 Hz to 100 khz in quasi-equilibrium conditions, by applying small biases to the nanotubes, V DS 6 50 mv. It should be noted that the small value of parasitic capacitance caused by the cables and the junction itself was not higher than C P = 200 pf, and therefore the roll-off frequency of our noise set-up is 160 khz (f roll off ¼ 1=ð2pC P ðr load jjr S ÞÞ 1=ð2pC P R load Þ160kHz). Here R S is the resistance of the samples under investigation. The experimental studies were performed in a vacuum of about 10 4 mbar over a large temperature range from 70 K to room temperature (T 298 K). where C G ¼ 2pL DS e r e 0 = lnð2t=rþ is the gate capacitance, L DS is the transistor channel length, V DS is the drain source voltage, e r is the dielectric constant of SiO 2, t is the thickness of the SiO 2 layer, and r is the radius of the CNT. Using e r = 3.9, V DS =30mV, L DS =6lm, and r = 0.5 nm, we obtain l 7000 cm 2 V 1 s 1 at room temperature. The value obtained considerably exceeds the values reported for conventional semiconductors. Then the total number of carriers, N, in the bias point with maximum transconductance can be calculated from [20]: j ¼ enle ) I S ¼ e N L DS S l V DS ; ð3þ L DS N ¼ L2 DS R N el 320; ð4þ where j is the current density, n is the concentration of carriers in the CNT, E is the electric field, I is the current, S is the radial section area of the CNT, and R N =1MX is the resistance of the CNT FET. The values of mobility and total number of carriers thus obtained will be used further for noise analysis. The investigated CNT FETs demonstrate p-type characteristic behavior. Drain current measured at fixed drain source voltage increases with increasing negative gate voltage. We registered the maximum of transconductance near zero gate voltage and the saturation regime at large negative gate voltage, higher than 1 V. Below we will show that CNT material determines transport and noise properties of the structures in this regime of large gate voltages. At the same time at small gate voltages the Schottky-barriermodulated regime is important for the transistor structure. We registered stable and reproducible behavior of the CNT FETs in both regimes: at small as well as at large gate voltages. We will start with detailed analysis of transport mechanisms in regime of small gate voltages and continue with characterization of the transport properties of the structures at large gate voltages. Transfer characteristics (Fig. 2), measured at small gate voltages and different temperatures demonstrate weak temperature dependence of the transconductances. Such behavior is characteristic of tunnel transport in the structure. The threshold voltage, V th, has a linear dependence in a wide temperature range (inset in Fig. 2). As in traditional metal oxide semiconductor FETs (MOSFETs), the temperature dependence of the surface inversion potential induces

4 CARBON 53 (2013) Fig. 2 Typical transfer characteristics of CNT FETs measured at different temperatures near the bias point with maximum transconductance. Inset: threshold voltage dependence on temperature. V DS =30mV. changes in the threshold voltage over temperature. This dependence can be expressed as: V th ðtþ ¼V th ðt 0 Þ a Vth ðt T 0 Þ; ð5þ where a Vth is the threshold voltage temperature coefficient. From a linear fit of the dependence presented in the inset of Fig. 2, we obtain a Vth ¼ 1:6 mv=k. It should be noted that such a value is typical of FET devices. For complementary metal oxide semiconductor (CMOS) devices, the coefficient ranges from 1 mv/k to 4 mv/k. It is calculated that this value for Macronix nonvolatile memory solutions (MXIC) 0.5 lm technology is about 1.22 mv/k and mv/k for n- channel and p-channel MOSFETs, respectively [21]. Thus CNT structures demonstrate behavior characteristic of FET devices, confirming the semiconductor nature of the conductivity of CNT material. Noise spectra were measured for each point of the transfer characteristics shown in Fig. 2, at different gate voltages and temperatures. Typical noise characteristics measured at different gate biases and at T = 80 K are shown in Fig. 3. An analysis of noise spectra demonstrates that they contain thermal, flicker and generation recombination noise components. It should be noted that GR noise components were not previously observed due to the relatively high level of flicker noise. The lowest horizontal value in Fig. 3a corresponds to the thermal noise component, which is calculated from S thermal V ¼ 4kTR eq, where k is the Boltzmann constant, T is the measured temperature, and R eq is the equivalent resistance equal to the parallel connected load, R load, and CNT, R S, resistances. It should be noted that at gate voltages higher than 0.7 V the normalized noise spectra have almost the same noise level (Fig. 3b, curves 5, 6). These results demonstrate that two different transport regimes can be analyzed separately in our CNT FET structures. The noise spectra measured at different temperatures are shown in Fig. 4. Samples demonstrate GR noise components in wide temperature and gate ranges. The characteristic frequency decreases as the sample cools down because of increases in the time constants of generation recombination processes inside the FET channel. The analysis of GR noise components reveals the presence of a few traps with different activation energies (see below). The normalized current noise spectral density of the flicker noise component as a function of temperature is shown in Fig. 5. The noise spectral density is almost independent of temperature at a fixed drain current or resistance of the CNT FET. This demonstrates that the number of carriers does not depend on temperature. The mobilities determined from measured characteristics (Fig. 2) are also almost independent of temperature. The equivalent input gate voltage spectral density, S U = S I /g m 2 (where g m is the transconductance of the FET obtained at each point of the transfer characteristics, S I is the current flicker noise spectral density), dependence on gate overdrive is shown at T = 200 K in Fig. 6a. Because of S u independence of gate voltages (at V G V th < 0.5), it can be concluded that the major source of flicker noise is related to gate and located in SiO 2 layer near the CNT channel [22]. In addition to the flicker noise, GR noise components with characteristic time, s, were clearly registered for all samples in the temperature range investigated. GR noise components allow us to determine the parameters of the traps. Using an Arrhenius plot (Fig. 6b), we estimated the energy level of the traps, E t, with respect to valence band (activation energy of the traps). Fig. 3 (a) Typical noise spectra measured at different gate biases V G (V): 1 0; 2 ( 0.1); 3 ( 0.2); 4 ( 0.6); 5 ( 0.7); 6 ( 1), T =80K,V DS = 30 mv. (b) The normalized current noise power spectra.

5 256 CARBON 53 (2013) the sample resistance of about 0.8 MOhm at a certain gate voltage. The activation energies obtained at different bias points (different gate voltages) of the CNT FET are shown in Table 1 for one of the samples. The trap energies for other samples are almost the same. These traps are located near the interface between the CNT and the SiO 2. Additional information about trap density can be obtained from the plateau of the GR noise. This component of the noise spectrum is described by [23]: f 0 S GR I ¼ 1 Dn 2 2 I 2 p N n 1 þðf=f 0 Þ ; ð6þ 2 Fig. 4 Typical noise characteristics measured at V G = 0.6 V and at different temperatures. Inset: normalized current noise power spectrum density as a function of frequency demonstrating the temperature behavior of a GR noise component. Fig. 5 Normalized current noise power spectral density of the flicker noise component dependence on temperature at f = 1 Hz measured for different resistances of the CNT FET controlled by the gate voltage V G ; V DS =30mV. The energies were found to be 290 ± 10 mev, 200 ± 10 mev and 45 ± 5 mev for the case of the bias point corresponding to where N is the total number of carriers, n is the concentration of carriers, f 0 and S GR I are the characteristic frequency and the current noise spectral density of the GR process, respectively. Here Dn 2 ¼ðn 1 þ n 1 t þ p 1 t Þ 1 ð7þ n t and p t is the concentrations of the electrons and holes on the trap levels, respectively. The temperature dependence of the normalized current noise power spectral density is shown in Fig. 7, where f 0 S GR I0 =I2 corresponds to f 0 S GR I =I 2 at zero frequency. Estimating the number of carriers from the equation N ¼ L 2 DS =elr N (at R N = X, l = 7000 cm 2 V 1 s 1 ), and using the maximum of f 0 S GR I0 =I2, we can calculate that Dn 2 =n 1, which is in good agreement with the theory of generation recombination noise [23]. In addition, this relation demonstrates that n is smaller than n t and p t taking into account Eq. (7). Therefore the concentration of the traps is much higher than the concentration of free carriers in CNT structures. These results confirm the correctness of the value determined for the total number of carriers N = Separation and analysis of two transport regimes in CNT structure A more precise analysis of the flicker noise component of the CNT FETs at higher gate voltages allowed us to recognize two different regions (Fig. 8a) separated by a resistance change of approximately R 600 kx (V G V th 0.6 V). The first is the nonlinear region of the transfer characteristic, whereas the Fig. 6 (a) Typical equivalent input gate voltage spectral density dependence on gate overdrive (V G V th ) measured for two CNT FETs (1 and 2) at V DS = 30 mv and T = 200 K. (b) Arrhenius plot shown for one of the samples. Dashed lines correspond to linear fitting, which allows us to calculate the energy levels of the traps.

6 CARBON 53 (2013) Table 1 Typical activation energies obtained for one of the samples in different bias points of the CNT FET. * indicates that the activation energy was difficult to estimate due to the overlapping GR noise components. R (MX) GR 1 (mev) GR 2 (mev) GR 3 (mev) 0.6 ± ± ± ± ± ± ± ± ± ± ± ± ± 5 75 ± 4 >1.3 * * 66 ± 4 barrier. In the case of a CNT FET, V b depends on the applied drain source voltage, V DS, and gate voltage, V G. At the constant value of V DS Eq. (8) can be written as: I D T 2 exp erv G kt ; ð9þ where r is a coefficient that shows how gate voltage influence the barriers. Using the Hooge model [14] and assuming that the Hooge parameter, a H, does not depend on gate bias, the following relation for flicker noise component can be obtained: fs Flicker I T 2 I 2 D exp erv G kt : ð10þ Fig. 7 Normalized noise power spectral density of the plateau of a GR noise component (measured in the bias point with maximum transconductance at V DS = 30 mv) as a function of inverse temperature. The dashed line shows the value corresponding to GR noise plateau maximum, at which N = 320 was calculated for Dn 2 =n ¼ 1 using Eq. (6). second region shows the saturation regime (Fig. 8b). Let us consider the reason for two regions in more detail. The work function of the Pd (deposited above a thin adhesion layer of Cr and used as a contact to CNT) has been reported to be in the range of ev [24] while the electrochemical affinity of the CNT is in the range of ev [25]. Therefore, it is expected that Pd metallization results in contacts with transparent barriers for carrier transport. We observed a decrease of the current with an increase of the gate voltage of the FETs that normally corresponds to the p-type behavior of the channel. The diameter of the nanotubes in the CNT FET devices has the value 1 nm corresponding to an energy band-gap in the range of ev [26 28]. At small values of the applied drain source voltage, Schottky barriers at the interface between CNT and metallization determine the peculiarities of transport and fluctuations in the system as is shown in Fig. 8d. By applying gate voltage, the width of the Schottky barriers changes thus controlling the whole current in the structure. The current through the Schottky barriers has an exponential behavior as a function of applied voltage [20]: I b T 2 exp ev b ; ð8þ kt where T is the temperature, e is the electron charge, k is the Boltzmann constant, and V b is the voltage applied to the At small gate voltages, CNT structures demonstrate exponential dependencies of noise values on gate voltage. The exponential behavior is also observed in the transfer characteristic as a function of temperature (inset to Fig. 8b). Because the flicker noise component also displays exponential behavior (Fig. 8c), one can conclude that the main noise source is due to the Schottky barriers in this range of gate voltages. The coefficient r can be calculated using Eq. (10). Fig. 9 shows a linear dependence of the slope er/kt as a function of inverse temperature for one of the measured samples. We obtained r values equal to (0.22 ± 0.02), (0.02 ± 0.01) and (0.050 ± 0.007) for three samples. Values of these coefficients reflect how strong the influence of the back gate voltage on the barriers is and correlate with current value in the CNT structure. In the case of lower coefficient r values, samples demonstrate higher drain current in the on state. The small values of the coefficient r obtained for some of the samples can be explained by the decreased influence of the contact areas on the transport properties of the CNT FETs. This is also confirmed by increased current in the transfer characteristics of the samples in the on state at high gate voltages. At large negative gate voltages drain current is almost independent on gate overdrive (Fig. 8b). The normalized current noise spectral density of the flicker noise component as a function of current and temperature is shown in Fig. 10. It should be emphasized that in the sub-threshold regime the noise is much higher compared to the channel regime. It is known that the relative level of flicker noise of different kinds of materials can be estimated according to the Hooge relation [14]: fs I I ¼ a H 2 N ; ð11þ

7 258 CARBON 53 (2013) Fig. 8 (a) Normalized current noise power spectral density of the flicker noise component as a function of gate overdrive measured at different temperatures, T (K): 70, 120, 140, 160, 180, 200, 260, 280, and 296; V DS =30mV,f = 1 Hz; (b) drain current dependence on gate overdrive voltage measured at different temperatures. Inset represents the exponential behavior near zero V G V th. (c) Logarithm of normalized current noise power spectral density multiplied by squared temperature as a function of gate overdrive voltage. (d) Schematic band diagram of the single-walled CNT FET at zero gate and drain voltages. Fig. 9 The dependence of the slope of the flicker noise components (dashed lines in Fig. 8c) on inverse temperature obtained for one of the samples with strong influence of the gate voltage on contact areas. V DS =30mV. where a H is the Hooge parameter, I is the current, f is the frequency, and N is the number of carriers. The substitution of the total number of carriers of N = 320 into Eq. (11) and using the value of effective mobility for the carriers calculated Fig. 10 Normalized current noise power spectral density of the flicker noise component dependence on drain current measured at different temperatures in sub-threshold (Schottky determined) and on state (channel determined) regimes of the CNT FET. Drain source bias was varied from 10 mv to 50 mv. above allows us to estimate the Hooge parameter as a H = This value is higher than the Hooge parameter

8 CARBON 53 (2013) value suggested for conventional semiconductors [14], which additionally demonstrates the contact-defined transport regime in the structure. At large negative gate voltages, the CNT FET current is almost independent of (V G V th ). Here the CNT FET is in the on state; therefore a relatively large current through the channel (Fig. 8b) and the very thin Schottky contacts reflect the regime of CNT-determined transport. It should be emphasized, that in this regime the current noise spectral density is one order of magnitude lower (Fig. 8a) compared to the density registered at small gate voltages and the Hooge parameter is found to be about The parameter value is comparable to typical values for conventional semiconductors and thus this additionally confirms the transport regime determined by CNT material properties. 5. Conclusion In summary, we analyzed transport properties of FETs based on individual semiconductor CNTs using noise spectroscopy. Two transport regimes were found and analyzed separately. It was demonstrated that the main source of flicker noise at small gate voltages is related to fluctuations of the Schottky barriers in the contact region of the carbon nanotube. In this case, a relatively high Hooge parameter was obtained (a H = ). At the same time, in the on state of the CNT FETs, we registered decreased noise level corresponding to the negligible influence of Schottky barrier regions on the conductivity of the FET structures. A Hooge parameter at least one order of magnitude lower is obtained in this case of pure channel-determined transport. In contrast to previously reported data on a high level of flicker noise, in our structures we revealed GR noise components due to decreased 1/f noise level. The analysis of Lorentzian-shaped noise components allows us to analyze traps influencing the noise and transport properties of CNT material. The concentration of free carriers in the CNT channel was found to be lower than trap concentrations. The number of carriers in the operation point of FETs with maximum transconductance was found to be about 320. The FETs demonstrate stable properties at various temperatures. The high values of the mobility obtained are characteristic of high-speed transport in the channel of the CNT FETs demonstrating good prospects for utilizing the unique properties of CNT materials. REFERENCES [1] Tans SJ, Verschueren ARM, Dekker C. Room-temperature transistor based on a single carbon nanotube. Nature 1998;393: [2] Heinze S, Tersoff J, Avouris P. Carbon nanotube electronics and optoelectronics. Lect Notes Phys 2005;680: [3] Martel R, Derycke V, Lavoie C, Appenzeller J, Chan KK, Tersoff J, et al. Ambipolar electrical transport in semiconducting single-wall carbon nanotubes. Phys Rev Lett 2001;87(25): [4] Bachtold A, Hadley P, Nakanishi T, Dekker C. Logic circuits with carbon nanotube transistors. Science 2001;294: [5] Freitag M, Radosavljevic M, Zhou Y, Johnson AT, Smith WF. Controlled creation of a carbon nanotube diode by a scanned gate. Appl Phys Lett 2001;79(20): [6] Heinze S, Tersoff J, Martel R, Derycke V, Appenzeller J, Avouris P. Carbon nanotubes as Schottky barrier transistors. Phys Rev Lett 2002;89(10): [7] Chen Z, Appenzeller J, Knoch J, Lin Y, Avouris P. The role of metal-nanotube contact in the performance of carbon nanotube field-effect transistors. Nano Lett 2005;5(7): [8] Franklin AD, Chen Z. Length scaling of carbon nanotube transistors. Nat Nanotechnol 2010;5(12): [9] Solomon PM. Contact resistance to a one-dimensional quasiballistic nanotube/wire. IEEE El Dev Lett 2011;32(3): [10] Nemec N, Tománek D, Cuniberti G. Contact dependence of carrier injection in carbon nanotubes: an ab initio study. Phys Rev Lett 2006;96(7): [11] Kim W, Javey A, Tu R, Cao J, Wang Q, Dai H. Electrical contacts to carbon nanotubes down to 1 nm in diameter. Appl Phys Lett 2005;87(17): [12] Collins PG, Fuhrer MS, Zettl A. 1/f noise in carbon nanotubes. Appl Phys Lett 2000;76(7): [13] Kim K, Jang D, Lee K, Kang H, Yu BY, Lee JI, et al. Influence of electrical contacts on the 1/f noise in individual multi-walled carbon nanotubes. Nanotechnology 2010;21: [14] Hooge FN. 1/f noise is no surface effect. Phys Lett A 1969;29(3): [15] Ishigami M, Chen JH, Williams ED, Tobias D, Chen YF, Fuhrer MS. Hooge s constant for carbon nanotube field effect transistors. Appl Phys Lett 2006;88: [16] Lin Y-M, Appenzeller J, Knoch J, Chen Z, Avouris P. Lowfrequency current fluctuations in individual semiconducting single-wall carbon nanotubes. Nano Lett 2006;6(5): [17] Behnam A, Biswas A, Bosman G, Ural A. Temperaturedependent transport and 1/f noise mechanisms in singlewalled carbon nanotube films. Phys Rev B 2010;81(12): [18] Tobias D, Ishigami M, Tselev A, Barbara P, Williams ED, Lobb CJ, et al. Origins of 1f noise in individual semiconducting carbon nanotube field-effect transistors. Phys Rev B 2008;77(3): [19] Danilchenko BA, Tripachko NA, Lev S, Petrychuk MV, Sydoruk VA, Sundqvist B, et al. 1/f noise and mechanisms of the conductivity in carbon nanotube bundles. Carbon 2011;49(15): [20] Sze SM. Physics of semiconductor devices. 2nd ed. New York: Wiley; [21] Uy KJ, Reyes-Abu PA, Chung WY. A high precision temperature insensitive current and voltage reference generator. World Acad Sci Eng Technol 2009;50: [22] Simoen E, Claes C. On the flicker noise in submicron silicon MOSFETs. Solid-State Electron 1999;43(5): [23] Lukyanchikova NB. Noise research in semiconductor physics. CRC Press; [24] Juutilainen A, Ahlskog M, Volodin A. Measurements of tunneling conduction to carbon nanotubes and its sensitivity to oxygen gas. Phys Rev B 2012;86: [25] Buonocore F, Trani F, Ninno D, Di Matteo A, Cantele G, et al. Ab initio calculations of electron affinity and ionization potential of carbon nanotubes. Nanotechnology 2008;19: [26] Dresselhaus MS, Dresselhaus G, Eklund PC. Science of fullerenes and carbon nanotubes. Academic Press; [27] Odom TW, Huang JL, Kim P, Lieber CM. Atomic structure and electronic properties of single-walled carbon nanotubes. Nature 1998;391:62 4. [28] Wildoer JWG, Venema LC, Rinzler AG, Smalley RE, Dekker C. Electronic structure of atomically resolved carbon nanotubes. Nature 1998;391:59 62.