Narrow Results

We report results of the experimental investigation of the low-frequency noise in graphene transistors. The graphene devices were measured in three-terminal configuration. The measurements revealed low flicker noise levels with the normalized noise spectral density close to 1/f (f is the frequency) and the Hooge parameter α_H ~10^-3. Both top-gate and back-gate devices were studied. The analysis of the noise spectral-density dependence on the gate biases helped us to elucidate the noise sources in these devices. We compared the noise performance of graphene devices with that of carbon nanotube devices. It was determined that graphene devices works better than carbon nanotube devices in terms of the low-frequency noise. The obtained results are important for graphene electronic, communication and sensor applications.

A modified charge fluctuation low-frequency noise model for an electrolyte-insulator-semiconductor (EIS) structure is developed. Physical processes in the semiconductor, insulator and electrolyte medium responsible for low-frequency charge fluctuation are discussed based on an electrical equivalent scheme for the EIS structure. Noise spectral density dependence on charge concentration fluctuation related to processes on the electrolyte-insulator, insulator-semiconductor interfaces and bulk semiconductor are analyzed.

In this work, a new, physically based model for the low-frequency noise is investigated by statistical simulations. The proposed model is based only on superposition of generation-recombination centers, and can predict the frequency-, current- and area-dependence of the low-frequency noise, as well as the area-dependence of the variation in the noise level. Measurements on Bipolar Junction Transistors (BJTs) are found to be in excellent agreement with the simulated results. For devices with large emitter areas A_E, the model predicts a spectral density S_In ~ 1/f. For devices with submicron A_E, S_In strongly deviates from a 1/f behavior, and several generation-recombination centers dominate the spectrum. However, the average spectrum <S_In>, calculated from several BJTs with identical A_E, has a frequency dependence ~ 1/f. The extracted areal trap density within the frequency range 1-10⁴ Hz is n_T = 3 × 10⁹cm^-2. The simulations show that the condition for observing g-r noise in the spectrum, strongly depends on the number of traps N_T, as well as the distribution of the corresponding energy level for the traps. The relative noise level is found to vary in a non-symmetrical way around < S_In>, especially for small A_E. For A_E < 0.1 μm², the model predicts a relative variation in the noise level below <S_In>, and above <S_In>. For A_E > 0.3 μm², the variation is found to be .

We report detailed investigations of low-frequency excess noise in Ga-polarity GaN thin films deposited by RF-plasma assisted molecular beam epitaxy. The noise properties of the GaN thin films deposited with and without the intermediate-temperature buffer layers (ITBL) are studied in detailed to examine the effects of the ITBL on the noise. Substantial reduction in the flicker noise levels are observed for samples grown on ITBLs with a Hooge parameter of 3×10^-4, which is believed to be the lowest, to date, reported for GaN material. At low-temperatures, Lorentzian bumps originating from the generation-recombination processes are observed. Detailed studies of the temperature dependencies of the voltage noise power spectra have led to the formulation of a model for the observed low-frequency fluctuations. The model stipulates that the phenomenon arises from the thermally activated trapping and detrapping of carriers. The process results in the correlated fluctuations in the carrier number and the Coulombic scattering rate. Quantitative computation shows that number fluctuation dominates in our samples. Numerical evaluation of the deep-levels indicates substantial reduction in the trap density for the Ga-polarity GaN films.

Low frequency noise (LFN) was measured in δ-doped GaAs structures in which the free carriers are confined to a 2-dimensional plane. Three samples grown at different temperatures, resulting in doping layers of a different thickness, are used to study the effects of quantum confinement on the LFN. We observed both 1/f noise and generation-recombination noise components. We find that a stronger quantum confinement results in a bigger Hall mobility and a lower magnitude of the 1/f noise.

The low frequency noise (LFN) behavior in 0.15 μm N-channel fully depleted SOI/MOS transistors is investigated, both in ohmic and saturation regimes. The extraction of the interface trap density is also carried out. Moreover, the existence of a moderate kink effect and its influence on LF noise in fully depleted devices are shown.

Hall sensors are used in a very wide range of applications. A very demanding one is electrical current measurement for metering purposes. In addition to high precision and stability, a sufficiently low noise level is required. Cost reduction through sensor integration with low-voltage/low-power electronics is also desirable. The purpose of this work is to investigate the possible use of SOI (Silicon On Insulator) technology for this integration. We have fabricated SOI Hall devices exploring the useful range of silicon layer thickness and doping level. We show that noise is influenced by the presence of LOCOS and p-n depletion zones near the edges of the active zones of the devices. A proper choice of SOI technological parameters and process flow leads to up to 18 dB reduction in Hall sensor noise level. This result can be extended to many categories of devices fabricated using SOI technology.

The current-voltage characteristics and the low-frequency noise spectra of p-type Si–Porous Si–Al diode like structures were investigated. Over 1 V forward biases a reasonable fit was obtained in the Fowler-Nordheim plot. Any attempts of accurately fitting the I-V characteristic by other known transport mechanisms failed. At lower biases, however, an additional current-component appears which shows a saturating character. This current component is ascribed to trap-assisted tunneling. The measured noise spectra show 1/f character; however the magnitude of the noise shows saturation with increasing biases instead of the usual case, where the noise power scales with I², or V². This finding is interpreted by a model of two parallel current paths. The noise arising from the smaller and saturating current determines the noise performance of the whole device.

Low-frequency excess noise was measured on laser-debonded HVPE-grown GaN films. While the Schottky MSM devices, fabricated on 5 μm-thick films, demonstrated reduction in the noise level, the ohmic MSM devices showed significant increase in the low-frequency noise. To account for the data we conducted two-dimensional simulation of the temperature of the GaN layer as a function of time and distance from the GaN/sapphire interface. The simulation results show that the temperature at the GaN/sapphire interface may rise up to 1400 K, whereas the maximum temperature at the GaN surface was about 400 K at the. Based on the experimental data and the simulation results, it is postulated that the illumination of the GaN sample by high-power excimer laser led to the decomposition of GaN at the GaN/sapphire interface resulting in the generation of localized states at the interface. The same process may have led to some thermal annealing effect at the GaN surface. To provide experimental proof of the hypothesis, detailed low-frequency noise measurement on ohmic MSM devices fabricated on 20 μm-thick GaN films were performed. The results indicate similar noise level for both the debonded and the control devices.

A model for the 1/f noise in large signal operation of linear passive one-port (e.g. carbon or polysilicon resistors) is given. Starting from the Hooge's formula, that holds when the component is operated under DC bias, it is shown that the noise current is simply proportional to the product between the conductivity fluctuation and the time-dependent large signal applied to the component. Otherwise stated, the 1/f noise exhibited by passive components in large signal operation arises from the intermodulation between the stochastic process "conductivity fluctuation" and the signal applied to the component. Detailed calculations of both autocorrelation function and power spectrum of the resulting noise current are given; different time dependence of the signal applied to the component are considered. Of particular relevance in practical applications is the case of sinusoidal signal plus an eventual DC bias; in this case the resulting noise current is a cyclostationary stochastic process, and its behaviour can be conveniently described by the cyclic autocorrelation functions or by the cyclic power spectra. The measured power spectra of the noise current of carbon resistors with DC bias and large periodic signal applied to them, are in good agreement with those calculated from the proposed model. The agreement of the measured power spectra with those predicted by the model allows to conclude that the physical origin of 1/f noise in passive components does not depend on the applied bias, DC or large signal alternating, since, in any case, the noise current is conveniently modeled by intermodulation (product) between the "conductivity fluctuation" stochastic process and the applied signal.

We report on the degradation of low-frequency excess noise in recessed gate AlGaN/GaN HEMTs due to hot-electron stressing. The I-V characteristics and the low-frequency noise power spectral densities, S_V(f), of the open circuit voltage fluctuations across the drain source terminal were characterized with the stress time. Based on these results, we observed that the overall low-frequency noise degradation process can be identified to occur in two distinct phases. In the first phase, devices initially show fluctuations in the noise properties around a relatively constant average value. Detailed characterizations of the gate-source bias, V_GS, dependence of S_V(f) at cryogenic temperatures indicate signature-patterns in the variations of S_V(f) as a function of V_GS. This is shown to arise from the modulation of the percolation paths of the carriers in the two-dimensional electron gas (2DEG). The onset of the second phase of degradation arises from the irreversible generation of interface states at the AlGaN/GaN hetero-interface.

Often the 1/f noise in MOSFETs is stated to be an ensemble of many RTS with different time constants. The majority of literature on 1/f noise is overlooking the contribution due to mobility fluctuations that are uncorrelated with number fluctuations. Here, we demonstrate that the so-called proofs for ΔN can also be obtained from the empirical relation. The following misunderstandings and controversial topics on the surface and bulk contributions to the low-frequency noise will be addressed: 1) 1/f and RTS noise can have different physical origins. An analysis in time domain shows that the low-frequency noise with RTS is nothing else than a superposition of a two level noise with a Lorentzian spectrum and a Gaussian noise with a pure 1/f spectrum and different bias dependency. 2) It is very unlikely that in a spectrum consisting of one strong two level RTS and a pure 1/f noise, the 1/f noise is a superposition of many RTS with different time constants. 3) The spreading in WLS_I /I² below a critical WL is not a proof for the ΔN origin. 4) The typical shape in the double log plot from sub threshold to strong inversion of S_I/I² versus I, is also not a proof for the ΔN origin.

Optical and electrical noises and correlation factor between optical and electrical fluctuations of nitride-based light emitting diodes (LEDs) have been investigated under forward bias. Their electrical, optical and noise characteristics were compared with ones of LEDs of other materials. LED noise characteristic changes during aging have been measured, too. It is found that optical and electrical noise spectra under forward bias for more reliable LEDs distinguish by lower l/f type fluctuations and Lorentzian type noise at higher frequencies. LEDs with intensive 1/f noise demonstrate shorter lifetime. It is shown that reason of LED degradation is related with defects presence in device structure. These defects can be formed during device fabrication or appear during operation. An analysis of LED current-voltage and electrical noise characteristics under forward and reverse bias has shown that LEDs with intensive 1/f electrical noise, large reverse current (low reverse breakdown voltage) and larger terminal voltage under forward bias distinguish by short lifetime.

A detailed study of photosensitivity and noise characteristics of ultrafast InGaAsP/InP avalanche photodiodes (APDs) with separate absorption, grading, charge and multiplication regions was carried out. Carrier multiplication and noise factors were evaluated. Noise origin in investigated APDs is 1/f, generation-recombination and shot noises. Different quality samples have been investigated and it is shown that noise characteristics well reflect APD quality problems. It is shown that low-frequency noise and excess shot noise characteristics are very sensitive to the APD quality problems and clear up physical processes in device structure. Noise characteristic analyses can be used for the APD quality problems revealing and optimal design development.

Hot-electron stressing on GaN-based light emitting diodes (LEDs) has been performed using a dc stressing current. Detailed characterizations on the degradations in the optoelectronic and low-frequency noise properties of the devices have been conducted. Experimental results on I-V, electroluminescence (EL) and the integrated light output of the devices exhibit significant degradations with the stress time. Investigations on the low-frequency excess noise of the devices show that the degradations of the device properties arise from the generation of interface states at the InGaN/GaN heterointerface due to hot-electron stressing. Over two orders of magnitude increase in the current noise power spectral density, S_I(f), is typically observed prior to the failure of the devices. Furthermore, it is shown that the initial rate of increase in S_I(f) resulting from a dc current stress time of 48 hours is strongly correlated to the lifetimes of the devices demonstrating that flicker noise measurement can be used as a diagnostic tool for hot-electron hardness of the device.

Resistive switching in aluminum-polymer diodes has been investigated by noise measurements. Quantitative criteria to characterize the diode states are: (i) Pristine state shows I ∝ V^m with m ≈ 6 at higher bias typical for tunneling. The resistance is very high, 1/f noise is very low, but the relative 1/f noise, fS_I/I² ≡ C_1/f is very high. (ii) Forming state is a time-dependent soft breakdown in the Al-oxide that results in random telegraph signal noise (RTS) with a Lorentzian spectrum or in multi-level resistive switching (MLS) with a 1/f^3/2 or 1/f-like spectrum. (iii) The H- or L-state shows I ∝ V^m with m = 1 for V < 1V and 3/2 < m < 2 for V > 1V. Deviations from ohmic behavior are explained by space charge limited current in the polymer. Reliable H- and L-states show pure 1/f noise, a resistance R that changes by at least a factor 30 and 1/f noise that follows the proportionality: C_1/f ∝ R with a proportionality factor αμ(cm²/Vs) of the same level as observed in metals, polymers and other semiconductors. C_1/f ∝ R is explained by switching of the number of homogeneous conducting paths in parallel. Deviations in C _1/f ∝ R are also explained. In the pristine state and even in the H-state, only a fraction of the device are is carrying current and switching seems to be at spots of the Al/Al₂O₃/polymer interface.

The low-frequency noise (LFN) in downscaled silicon transistors has become prominently large, and it occurs as a limiting factor for diverse applications. Considerable interest is paid to the “slow” (as compared to the operating frequency of the devices) noise. Therefore, we address the trends for LFN from an extensive analysis of data from many publications over a very long period. The impact of LFN on high-frequency device performance, the penalties associated with using composite materials, and unsolved issues are also discussed.

This paper presents analyses of the 3D acoustic fields generated by motorcycles at very low frequencies ($<150$ Hz) at the idle speed and during sudden acceleration. Diagnosis and analyses of sound sources at low frequencies have always been a significant challenge because the directivity of low-frequency sound is very poor. To date, there are no research papers and/or reports that have demonstrated low-frequency sources localization and radiation patterns of any kind at high spatial resolution in 3D space. This study shows that by using sound viewer technologies, which include the passive sonic detection and ranging (SODAR), the Helmholtz equation least squares (HELS) method, advanced signal processing, denoising, etc., the locations of sound sources and visualization of the sound fields can be determined with high spatial resolution, even at frequencies below 150 Hz. In particular, the HELS method allows for reconstructing all the acoustic quantities, including the acoustic pressure, time-averaged acoustic intensity, time-averaged acoustic power on the source surfaces and in 3D space. The hardware needed consists of a 3D array with six free-field precision microphones with pre-amplifiers, a miniature wide-angle camera, an eight-channel digital signal processor and a laptop computer.

Often the 1/f noise in MOSFETs is stated to be an ensemble of many RTS with different time constants. The majority of literature on 1/f noise is overlooking the contribution due to mobility fluctuations that are uncorrelated with number fluctuations. Here, we demonstrate that the so-called proofs for ΔN can also be obtained from the empirical relation. The following misunderstandings and controversial topics on the surface and bulk contributions to the low-frequency noise will be addressed: 1) 1/f and RTS noise can have different physical origins. An analysis in time domain shows that the low-frequency noise with RTS is nothing else than a superposition of a two level noise with a Lorentzian spectrum and a Gaussian noise with a pure 1/f spectrum and different bias dependency. 2) It is very unlikely that in a spectrum consisting of one strong two level RTS and a pure 1/f noise, the 1/f noise is a superposition of many RTS with different time constants. 3) The spreading in WLS_I /I² below a critical WL is not a proof for the ΔN origin. 4) The typical shape in the double log plot from sub threshold to strong inversion of S_I /I² versus I, is also not a proof for the ΔN origin.

We report results of the experimental investigation of the low-frequency noise in graphene transistors. The graphene devices were measured in three-terminal configuration. The measurements revealed low flicker noise levels with the normalized noise spectral density close to 1/f (f is the frequency) and the Hooge parameter α_H ~10⁻³. Both top-gate and back-gate devices were studied. The analysis of the noise spectral-density dependence on the gate biases helped us to elucidate the noise sources in these devices. We compared the noise performance of graphene devices with that of carbon nanotube devices. It was determined that graphene devices works better than carbon nanotube devices in terms of the low-frequency noise. The obtained results are important for graphene electronic, communication and sensor applications.