Defect Pool Model Research Articles

Studies of defect states in the interface between SiO 2 and a-Si:H films

In this paper, the results of studies on the interface properties in the hydrogenated amorphous silicon thin-film transistors with silicon oxide gate insulators are presented by means of the bias-annealing method. The transfer characteristics before and after bias-annealing are measured. Using a computer program, the field-effect density-of-states are calculated from the transfer characteristics. The results show that the density of states distribution is determined by the Fermi energy during thermal equilibration. These experimental results are explained by the defect-pool model.

Understanding of Thin Film Transistor Electrical Measurements in the Light of the Defect Pool Model

ABSTRACTRecently we described a new computer program which used a defect pool model of the density of states in a thin film transistor. With this model we modelled the effects of thermal bias annealing on the field effect characteristics of amorphous silicon thin film transistors, and used these results to determine key defect pool parameters. Here we use this program to provide new insight into other features of thin film transistor properties, such as activation energies of conduction as a function of gate bias, details of the subthreshold slopes, Movement of the apparent flat band voltages, and contact effects including current crowding. These effects are all successfully modelled and explained with our first principles defect pool Model, and compared with experimental results on a variety of material systems. In particular we show how measurements of the apparent flat band voltage produce potentially misleading results, and explain the inconsistency between observations of much less current crowding in contacts compared to predictions based on observed bulk defect densities and conventional Models.

A Unified Description of the Density of States for Hydrogen and Electrons in Amorphous Silicon

ABSTRACTWe derive the hydrogen density of states for hydrogenated Amorphous silicon (a-Si:H), which is completely consistent with the electronic density of states for the defect pool model. If silicon dangling bond energies are distributed in energy, as in a defect pool model, then the hydrogen density of states becomes quite complex, with the hydrogen binding energy dependent on the Fermi level and dangling bond transition energy. We demonstrate that the electronic density of states for dangling bonds is almost identical to our previous defect pool model, while the hydrogen density of states can account for the results of hydrogenation-dehydrogenation experiments. The effective hydrogen correlation energy is variable, being negative for most hydrogen binding sites, but positive for most defect sites.

Defect pool model in amorphous silicon

Defect pool model in amorphous silicon

The Kinetics of Removal of Defects in a-Si:H TFTs Made with PECVD Oxide Gate Insulator Material

ABSTRACTAMorphous silicon thin film transistors may be manufactured using a variety of gate insulator Materials. The sign and magnitude of the fixed charge, Qf, in the gate insulator material determines the as-deposited density of states in the a-Si:H as predicted by the defect pool Model [1]. For the case of PECVD nitride Qf is positive giving rise to a defect peak in the lower part of the bandgap. In thermal oxide, which has small negative Qf the states are in the upper part of the bandgap. Using the technique of low temperature bias stress or high temperature bias anneal it is possible to create further defects in the bandgap on top of those already present, facilitating the study of upper (Dh) and lower (De) gap states. In a previous work we studied the removal kinetics of such defects using TSDR (Thermally Stimulated Defect Removal) in both PECVD nitride and thermal oxide devices [2]. Here we repeat the experiments on thin film transistors made with PECVD helium diluted oxide as the gate insulator Material. PECVD helium diluted oxide is interesting as its Qf May be varied during Manufacture. Thus TFTs with characteristics similar to either thermal oxide or PECVD nitride can be obtained.

Density of Deep Bandgap States in Amorphous Silicon From the Temperature Dependence of Thin Film Transistor Current

ABSTRACTWe have measured the current-voltage characteristics of amorphous silicon thin film transistors (a-Si TFTs) over a wide range of temperatures (20 to 160°C) and determined the activation energy of the channel current as a function of gate bias with emphasis on the leakage current and subthreshold regimes. We propose a new method for estimating the density of localized states (DOS) from the dependence of the derivative of activation energy with respect to gate bias. This differential technique does not require knowledge of the flat-band voltage (VFB) and does not incorporate integration over gate bias. Using this Method, we have characterized the density of localized states with energies in the range 0.15–1.2 eV from the bottom of the conduction band and have found a wide peak in the DOS in the range of 0.8–0.95 eV below the conduction band. We have also observed that the DOS peak in the lower half of the bandgap increases in magnitude and shifts towards the conduction band as a result of thermal and bias stress. We also measured an overall increase in the DOS in the upper half of the energy gap and an additional peak, centered at 0.2 eV below the conduction band, which appear due to the applied stress. These results are in qualitative agreement with the defect pool Model [1,2].

The Interpretation of the Constant Photocurrent Method in Terms of Deep Defect Density of States in A-SI.H

ABSTRACTThe constant photocurrent Method (CPM) is often used to measure the sub-bandgap absorption for the determination of the defect density. However, the absolute value of the derived defect density depends on the method of data analysis and the calibration factor. Normally the calibration factor is obtained from electron spin resonance (ESR) but the defect pool model gives rise to doubt whether ESR detects the same defects as CPm. Therefore, we propose combined total-yield photoelectron spectroscopy (TYPES) and CPM Measurements on n-type a-Si:H to determine the calibration factor. Furthermore, we calculate CPM spectra by extending an approach to simulate photoconductivity, taking into account the full set of optical transitions, and compare the results with standard evaluation Methods.

The defect-pool model and charged defects in amorphous silicon

A generalized concept of chemical equilibration including electronic occupation statistics is presented. The model applies to both equilibrium and metastable defect densities and distributions under general non-equilibrium conditions. In undoped low-defect a-Si:H, the model predicts a charged-to-neutral defect ration of ∼3–7 and a decrease of this ratio to ∼1 by light soaking, in agreement with data from subgap absorption, ESR and LESR.

Defect-pool model parameters for amorphous silicon derived from field-effect measurements

Field-effect conductance measurements are analysed using a new defect-pool model and energy-dependent density of states for amorphous silicon is derived. We determine the key defect-pool parameters and in particular we find that the energy separation (Δ) between the D − states in n-type a-Si:H and the D + states in p-type a-Si:H is 0.44 eV. This result implies that the equilibrium density of states in bulkintrinsic amorphous silicon must contain about four times more charged defects than neutral defects.

Physics of a-Si:H switching diodes

Amorphous silicon diodes can be used as simple, two-terminal switches in active matrix addressing of display and image sensor arrays. These switching diodes provide an alternative to transistors with a simpler technology and fewer pixel connections. This paper discusses different addressing schemes in which switching diodes can be used and it reviews the device physics for these applications. In particular, it addresses the mechanisms of the dark steady-state I–V characteristics and the transient response of the diodes under typical operation conditions. The discussion gives a qualitative approach to the device physics, which is supported by computer modelling that uses complete expressions for occupancy, charge density and recombination. The defect density distribution is calculated from first principles using the defect-pool model.

Field-effect conductance in amorphous silicon thin-film transistors with a defect pool density of states

A new computer program to analyze field-effect conductance measurements has been developed. In this program a defect pool model, where the equilibrium density of state is determined by the Fermi level, has been incorporated. Transistors with finite band bending, due to fixed charge in the insulator, will therefore have a density of states that is spatially inhomogeneous. The inhomogeneous density of states means that the subthreshold slope of a device is not always controlled by the density of states near the interface, but can become dominated by the bulk density of states, contrary to simpler models. Both electron and hole branches are modeled simultaneously and self-consistently with no assumptions made about the flatband voltage. Indeed, it is demonstrated that there is no flatband voltage in a transistor with an inhomogeneous density of state; however, a true flatband voltage can be achieved by a process of thermal bias annealing. Finite thickness effects and defect correlation energies are taken into account. The program is used to model the characteristics of thin-film transistors both before and after thermal bias annealing. The model fits the experimental results well, even though there are few free parameters. A key result is that the defect pool parameter Δ is determined, which is the energy separation of the D0/− transition observed in n-type material and the D+/0 transition in p-type material. It is found that Δ=0.44 eV for all devices fitted, which suggests this a fundamental parameter of a-Si:H. This value of Δ implies that there are about four times as many charged as neutral devices in bulk intrinsic amorphous silicon.

Defect distribution in a-Si:H-pin solar cells before and after degradation

Amorphous silicon pin solar cells suffer from a strong performance degradation due to light soaking. It is widely accepted that the microscopic cause of this effect is the creation of metastable defects. However, it is not yet clear whether the additional defects are created predominantly in the bulk of the i-layer or at the pi- and in-interface. Numerical modeling can be a very significant means to distinguish between these opposite assumptions, but absolutely needs an optimization of its parameters in order to produce reliable results. After doing this first step very carefully, we have modeled the effect of different spatial defect distributions in a series of model solar cells, which are compared with state of the art pin solar cells with varying i-layer thicknesses. The best coincidence for the annealed state is found if the defect density is computed selfconsistently with the varying position of the fermi level in analogy to a simplified defect pool model. We do not introduce an energy variance of the defects themselves. Furthermore, we find clear evidence that the light induced degradation leads to a homogeneous rise of the defect density in the bulk of the i-layer and not only at the pi- (and in-)interface.

Kinetics and steady-states of light-induced defects in a-Si:H prepared by three different techniques

Three different preparation techniques for hydrogenated amorphous silicon (a-Si:H), (1) the conventional plasma CVD method, (2) post-deposition annealing of a-Si:H prepared at a low deposition temperature, and (3) helium-dilution method, were employed to obtain a-Si:H with different microscopic/mesoscopic structures, of which light-induced metastability was investigated. Differences among the samples appeared in the steady-state defect density under illumination, its dependence on illumination temperature, and in thermal and light-induced annealing behaviors. They were interpreted in the framework of a defect-pool model, where the distributions of defect annealing energies depend on the film structure.

Improved defect-pool model for charged defects in amorphous silicon.

We have developed an improved defect-pool model for the calculation of the density of dangling-bond states in amorphous silicon. The results of this improved defect-pool model are contrasted with earlier work, from which we have eliminated some errors and approximations. We show that the calculated energy-dependent density of states depends on the specific microscopic reaction involving hydrogen, in contrast to previous conclusions. We calculate the bulk density ofstates, using the best input parameters drawn from experiment, and conclude that the best agreement with experimental results is found for a rather wide defect pool and for a microscopic model where two Si-H bonds break for every weak bond converted into two dangling bonds

Evidence for the defect-pool model from induced recombination level shifts in undoped a-Si:H.

The measured dependence of the mobility-lifetime products for electrons and holes, on the relative position of the dark Fermi level ${\mathit{E}}_{\mathit{F}}$ in undoped hydrogenated amorphous silicon, a-Si:H, is reported here. The corresponding measurements, which also included the ${\mathit{E}}_{\mathit{F}}$ dependencies of the light-intensity exponents \ensuremath{\gamma} and S, became feasible due to the use of a combination of metal-oxide-semiconductor and photocarrier grating configurations. We found that the experimental results of the S(${\mathit{E}}_{\mathit{F}}$) dependence are very helpful in determining which of the models proposed for the recombination-center distribution in the mobility gap of a-Si:H is the most appropriate one.

The role of the gate insulator in the defect pool model for hydrogenated amorphous silicon thin film transistor characteristics

We demonstrate that differing transistor characteristics in the most important material systems can be explained by the defect pool model applied to the defects near the interface of hydrogenated amorphous silicon thin film transistors. Gate dielectrics used include plasma deposited silicon nitrides, plasma deposited silicon oxides, and thermally grown silicon oxides. The most important property of the gate dielectric is not the chemical composition but the fixed charge. In particular, as-deposited plasma deposited silicon oxide transistors can be made with similar properties to plasma deposited silicon nitride transistors or thermal silicon oxide transistors, by varying the fixed charge. After correcting the effects of the fixed charge variation, some differences still exist between the interface qualities. We introduce the parameter Ndb(min), i.e., the minimum density of dangling bonds (cm−2), which is a measure of interface quality independent of the fixed charge of the gate insulator. We propose that the variations in Ndb(min) are due to differences in disorder, perhaps caused by interface strain, leading to slight variations (∼5 meV) of the valence-band tail slope.

Defect chemical potential and the density of states in amorphous silicon.

We derive a general expression for the defect chemical potential of amphoteric silicon dangling bond states in amorphous silicon. We show that a single expression is valid, irrespective of the charge state of the dangling bond. We apply this result in a defect pool model for the formation of dangling bonds by the breaking of weak silicon-silicon bonds, with and without hydrogen motion. We calculate the resulting energy dependent density of states.

Dependence of the Saturation of Light-Induced Defect Density in a-Si:H on Temperature and Light Intensity

We studied the temperature and intensity dependence of the saturated density of light-induced defects (Nsat) in hydrogenated amorphous silicon (a-Si:H), established by high-intensity Kr+ laser illumination. The saturation value is insensitive to temperature below about 90°C. Above 90°C, Nsat drops with increasing temperature. This behavior can be explained within the defect pool model by a limited number of defect sites coupled with the concept of defect equilibrium. The experimental data suggest that the tail states do not direcily affect Nsat in device-quality a-Si:H.

Defect pool in amorphous-silicon thin-film transistors.

Amorphous-silicon thin-film transistors show a threshold voltage shift when subjected to prolonged bias stress. For transistors made with silicon oxide as the gate dielectric, the threshold shift induced under positive bias is due to the creation of dangling-bond states in the a-Si:H at low energy (${\mathit{D}}_{\mathit{e}}$ states). The threshold shift induced by negative bias stress is due to the creation of dangling-bond states at a higher energy (${\mathit{D}}_{\mathit{h}}$ states). In transistors made with silicon nitride as the gate dielectric, positive bias stress causes an increase in the density of ${\mathit{D}}_{\mathit{e}}$ states, but negative bias stress causes mainly a reduction in the density of ${\mathit{D}}_{\mathit{e}}$ states. Positive bias annealing of both oxide and nitride transistors leads to an increase in the density of ${\mathit{D}}_{\mathit{e}}$ states and a reduction in the density of ${\mathit{D}}_{\mathit{h}}$ states. Negative bias annealing leads to a reduction in the density of ${\mathit{D}}_{\mathit{e}}$ states and an increase in the density of ${\mathit{D}}_{\mathit{h}}$ states. The magnitude of each change depends on the initial Fermi-level position, which is the main difference between our oxide and nitride transistors. The results are explained by a defect-pool model for the dangling-bond states in a-Si:H. Dangling bonds are formed by a chemical equilibration process, resulting in the formation of dangling bonds in each of the possible charge states. This leads to a density of states in a-Si:H consisting of coexisting components formed as negatively charged dangling bonds (${\mathit{D}}_{\mathit{e}}$ states), positively charged dangling bonds (${\mathit{D}}_{\mathit{h}}$ states), and neutral dangling bonds (${\mathit{D}}_{0}$ states).Fitting the calculated density of states to the experimental results determined from the transistor characteristics leads to the conclusion that the density of ${\mathit{D}}_{\mathit{e}}$ and ${\mathit{D}}_{\mathit{h}}$ states dominates over the ${\mathit{D}}_{0}$ states, for all Fermi-level positions. We therefore conclude that there must be substantial densities of charged dangling bonds, even in undoped a-Si:H. Because of the wide energy separation of the ${\mathit{D}}_{\mathit{e}}$,${\mathit{D}}_{0}$, and ${\mathit{D}}_{\mathit{h}}$ states, virtually all the states in the lower part of the gap are ${\mathit{D}}_{\mathit{e}}$ states and all the states in the upper part of the gap are ${\mathit{D}}_{\mathit{h}}$ states. Raising the Fermi level increases the density of ${\mathit{D}}_{\mathit{e}}$ states and lowers the density of ${\mathit{D}}_{\mathit{h}}$ states, but the density of ${\mathit{D}}_{0}$ states remains the same. Lowering the Fermi level increases the density of ${\mathit{D}}_{\mathit{h}}$ states and reduces the density of ${\mathit{D}}_{\mathit{e}}$ states. Bias stress of nitride transistors, at much higher fields, leads to larger threshold voltage shifts, due to charge trapping in the nitride. Subsequent annealing leads to a new zero-bias thermal-equilibrium density of states. Transistor characteristics can be optimized in this way.

Bias-stress-induced creation and removal of dangling-bond states in amorphous silicon thin-film transistors

We report on the threshold voltage shift in amorphous silicon thin-film transistors, subjected to a gate bias for a prolonged period of time (bias stress). For transistors made with a silicon nitride gate insulator, the threshold voltage shift for low positive bias is due to dangling-bond-state creation in the amorphous silicon layer. For low negative bias, the threshold voltage shift is due to the bias-stress-induced removal of dangling-bond states. These results are contrasted with previously published results for oxide transistors, but both results are consistent with a defect pool model for the dangling-bond states. The difference for oxide and nitride transistors is due to a different zero-bias Fermi energy position at the interface. For nitride transistors at much larger applied bias, the dominant mechanism changes and the threshold voltage shift is dominated by charge trapping in the gate dielectric. This is found for both large negative and large positive bias.