Prediction of plasma etching using a randomized generalized regression neural network

A new empirical technique to construct predictive models of plasma etch processes is presented. This was accomplished by combining a generalized regression neural network (GRNN) and a random generator (RG). The RG played a critical role to control neuron spreads in the pattern layer. The proposed R-GRNN was evaluated with experimental plasma etch data. The etching of silica thin films was characterized by a 2³ full factorial experiment. The etch responses examined include aluminium etch rate, silica etch rate, profile angle, and DC bias. Additional test data were prepared to evaluate model appropriateness. Compared to conventional GRNN, the R-GRNN demonstrated much improved predictions of more than 40% for all etch responses. This was illustrated over statistical regression models. As a result, the proposed R-GRNN is an effective way to considerably improve the predictive ability of conventional GRNN.     

In-situ monitoring of plasma equipment using spectrophotometric colorimetry and neural network

Plasma plays a critical role in fabricating thin films for manufacturing electronic devices. To improve equipment throughput and device yield, plasma should be strictly monitored. A new technique is presented to monitor process plasma. This is accomplished by using a spectrophotometer and a neural network. The spectrometer was used to calculate color chromaticity coordinates and color temperature. Their sensitivity was examined as a function of radio frequency source and bias powers. All the variables linearly increased or decreased for process-induced faults. For an in-situ monitoring, the color variables were modeled by using an auto-correlated type of a time-series neural network and model prediction was applied to a CUSUM control chart. Unlike conventional models, the neural network trained with a reference pattern was utilized for fault detection. The constructed models demonstrated prediction errors ranging from 0.018 to 0.027%. A presented model-based CUSUM yielded improved fault sensitivity over a raw sensor data-based one. The improvement was more drastic for the bias power-induced faults. The proven high sensitivity can be effectively applied to monitor plasma equipment in real-time.     

Modelling urban air quality using artificial neural network

This paper describes the development of artificial neural network-based vehicular exhaust emission models for predicting 8-h average carbon monoxide concentrations at two air quality control regions (AQCRs) in the city of Delhi, India, viz. a typical traffic intersection (AQCR1) and a typical arterial road (AQCR2). Maximum of ten meteorological and six traffic characteristic variables have been used in the models formulation. Three scenarios were considered—considering both meteorological and traffic characteristics input parameters; only meteorological inputs; and only traffic characteristics input data. The performance of all the developed models was evaluated on the basis of index of agreement (d) and other statistical parameters, viz. the mean and the deviations of the observed and predicted concentrations, mean bias error, mean square error, systematic and unsystematic root mean square error, coefficient of determination and linear best fit constant and gradient (Willmott in B Am Meteorol Soc 63:1309, 1982). The forecast performance of the developed models, with meteorological and traffic characteristics (d=0.78 for AQCR1 and d=0.69 for AQCR2) and with only meteorological inputs (d=0.77 for AQCR1 and d=0.67 for AQCR2), were comparable with the measured data.     

Use of a neural network to link AFM etch patterns to plasma parameters

Plasmas play a critical role in depositing thin films or etching fine patterns while manufacturing integrated circuits. A new model for plasma diagnosis is presented. This was accomplished by linking atomic force microscopy (AFM) to plasma parameters using a neural network. Experimental AFM data were collected during the etching of silicon oxynitride films in C₂F₆ inductively coupled plasma. Surface roughness of etched patterns was characterized by means of discrete wavelet transformation. This led to the construction of three vertical (type I), diagonal (type II), and horizontal (type III) wavelet coefficient-based models. The performance of diagnosis models was evaluated in terms of the prediction and recognition accuracies. Both accuracies were optimized as a function of the number of hidden neurons. Comparisons revealed that the type I model yielded the largest recognition and the smallest prediction error. This was demonstrated even under stricter monitoring conditions. More improved diagnosis is expected by enhancing AFM resolution.     

Low resistivity contacts to plasma etched Mg-doped GaN using very low power inductively coupled plasma etching

The electrical characteristics of contacts to p-type GaN etched using inductively coupled plasma-reactive ion etching at low radio-frequency (RF) and inductively coupled plasma (ICP) powers were studied via current-voltage characteristics and transfer length method analysis. Although the quality of ohmic contacts to p-GaN is generally degraded by plasma etching, the results indicate that the specific contact resistance improves significantly both for the same etch depth and for the same etch time when the power is reduced from RF/ICP power of 75/150 W to very low RF/ICP power of 20/40 W. The use of a two step approach using a combination of higher power (75/150 W) followed by very low power (20/40 W) etches on the same sample also show improvement compared to the standard etch. Finally, etching with low power to a shallow etch depth of 60 nm achieved a very low specific contact resistance of 0.035 Ωcm², while a deeper etch of more than 500 nm resulted in a contact resistance of 0.153 Ωcm², both lower than any previously published values for ohmic contacts made to etched p-GaN.     

Prediction of etching results and etching stabilization by applying principal component regression to emission spectra during in-situ cleaning

A method to predict etching results by analyzing plasma emission spectra during in-situ cleaning was investigated, where the plasma emission spectra indicate the surface condition of etching reactor walls. Plasma-wall interaction was evaluated by using both principal component regression of plasma emission spectra and attenuated-total-reflection Fourier-transform infrared spectroscopy. We found that differences in the amount of silicon oxide deposition on the reactor wall affected radical composition in the plasma during in-situ cleaning and consequently affected the etching results. Etching result predictions using the plasma spectra corresponded very well to the etching result measurements, which are used to improve etching stability.     

Microtrenching geometry of 6H-SiC plasma etching

Single-crystal 6H-SiC films were etched in a SF₆/O₂ inductively coupled plasma. Microtrenching which occurred in most experiments at the feet of the profile sidewall was characterized in terms of maximum depth and width. Each characteristic was examined as a function of the process parameters, including ICP coil power, bias voltage, and the O₂ percentage. Experimental results showed that microtrench is caused by the addition of O₂. Apart from the etch mechanisms, relationships between microtrenching and profile angle were also identified. In most cases, the depth and width variations were strongly dependent on the profile angle variation. The statistical experimental design of the process parameters showed that the percentage of O₂ was identified as the most important parameter. The addition of O₂ has influence on the effect of microtrench due to the formation of a SiF_xO_y layer, which have a greater tendency to charge than SiC.     

Microtrench depth and width of SiON plasma etching

Silicon oxynitride films were etched in a C₂F₆ inductively coupled plasma. In all experiments, microtrenching occurred at the feet of the profile sidewall. The microtrenching was characterized in terms of maximum depth and width. Each characteristic was examined as a function of the process parameters, including radiofrequency source power, bias power, pressure, and C₂F₆ flow rate. Apart from the etch mechanisms, relationships between microtrenching and profile angle were also identified. Profile angle variation played an important role in understanding depth variation. The width of microtrenching increased with increasing the source or bias power. In contrast, increasing the C₂F₆ flow rate decreased the width. Effect of process parameters on microtrenching at various plasma conditions was characterized by using a statistical experimental design. Smaller depths and widths were obtained at lower source and bias powers. The main effect analysis revealed that the bias power had a considerable impact on both characteristics.     

Hydrogen bromide plasma-copper reaction in a new copper etching process

The hydrogen bromide (HBr) plasma-copper reaction, which is the base of a new copper etching process, has been studied. A high etch rate has been obtained with the reaction. Influences of process parameters, such as plasma exposure time, pressure, plasma power and substrate temperature, to the reaction process were explored. In addition to the reaction rate, we also investigated the relationship between the morphology and structure of the reaction product layer and the process condition. The results show that both the plasma-phase chemistry and the ion bombardment energy are important to the reaction. Mechanism for the HBr plasma-Cu reaction process was compared with that for the HCl or Cl₂ plasma-Cu reaction. The reaction product was characterized with X-ray photoemission spectroscopy and X-ray diffraction. These results are critical to the understanding of this new copper dry etching process.     

Surface roughness of silicon carbide etching in a NF3 inductively coupled plasma

Silicon carbide was etched in a NF₃/CH₄ inductively coupled plasma. Surface roughness measured by atomic force microscopy was investigated as a function of process parameters. Both etch rate and dc bias were correlated to the surface roughness. To optimize the surface roughness, a 2⁴ full factorial experiment was conducted for 700-900 W source power, 50-150 W bias power, 0.80-1.60 Pa, and 20-100% NF₃ percentage. Main effect analysis revealed that the surface roughness is the most strongly affected by the bias power. For variations in the bias power or NF₃ percentage, decrease in the surface roughness was observed only as positive variations in the etch rate and dc bias are considerably large. The surface roughness with the pressure was chemically dominated as illustrated by its inverse relationship with the dc bias. For the variations in the NF₃ percentage, the radical variation was estimated to play a more dominant role. The smoothest surface roughness of 0.312 nm was obtained at 700 W source power, 150 W bias power, 1.60 Pa pressure, and 100% NF₃ percentage.     

Basic study of a glass substrate in dry etching system

Electrostatic Chucks (ESCs) in the dry etchers for the TFT-LCD fabrications have been investigated briefly. The behaviors of glass encountering electric fields in the presence of plasma were studied. In some conditions, it has been shown that the electric force and the pressure by He backside cooling acting on the glass might be larger than the gravitational force. With a simple model, the stability condition of glass substrate was obtained.     

Oxygen plasma etching of silver-incorporated diamond-like carbon films

Diamond-like carbon (DLC) film as a solid lubricant coating represents an important area of investigation related to space devices. The environment for such devices involves high vacuum and high concentration of atomic oxygen. The purpose of this paper is to study the behavior of silver-incorporated DLC thin films against oxygen plasma etching. Silver nanoparticles were produced through an electrochemical process and incorporated into DLC bulk during the deposition process using plasma enhanced chemical vapor deposition technique. The presence of silver does not affect significantly DLC quality and reduces by more than 50% the oxygen plasma etching. Our results demonstrated that silver nanoparticles protect DLC films against etching process, which may increase their lifetime in low earth orbit environment.     

Black SiC formation induced by Si overlayer deposition and subsequent plasma etching

Black SiC formation by plasma etching with SF₆/O₂ chemistry is reported. Black SiC was produced by depositing Si overlayer on SiC and then etching the Si/SiC stack sequentially, thus replicating the black Si morphology to SiC. Black SiC is obtained with almost zero reflectance over the wavelengths from 300 nm to 1050 nm. Thicker Si film was advantageous, and it was important to optimize the etch condition considering both the black Si morphology and the flattening effect of SiC.     

Response probability estimation for randomly excited quasi-linear systems using a neural network approach

A quasi-linear system is referred to as a system linear in properties and subjected to multiplicative random excitations appearing also in the linear terms. It is known that exact solutions for the stationary moments can be obtained analytically for such a quasi-linear system if the excitations are Gaussian white noises. However, the exact response probability, which is non-Gaussian, is not obtainable analytically. In this paper, a neural network approach is proposed to evaluate the stationary response probability for quasi-linear systems under both additive and multiplicative excitations of Gaussian white noises based on the obtained exact statistical moments. Numerical examples show that the procedure yields accurate results if an appropriate form is assumed for the probability density function. The accuracy of the results is substantiated by comparing them with those obtained from Monte Carlo simulations.     

Dry etching of AlN films using the plasma generated by fluoride

The plasma produced by the mixture of fluoride and argon (SF₆/Ar) was applied for the dry etching of AlN films. Very high etching rate up to 140 nm/min have been observed. The effects of the bias voltage and the plasma component on the etching results were investigated. It shows that AlN can be effectively etched by the plasma with the moderate SF₆ concentration and the etching rate varies linearly with the bias voltage. The FTIR spectra confirm that AlF₃ is formed due to the chemical reaction of Al and F atoms. The mechanism of AlN etching in F-based plasma is probably a combination between physical sputtering and chemical etching and can be briefly outlined: (i) F⁻ ions reacts with Al atoms to form low volatile product AlF₃ and passivate the surface, and (ii) at the same time the Ar⁺ ions sputter the reaction product from the surface and keep it fluoride free to initiate further reaction. AlF₃ formed on the patterned sidewall play a passivation role during the etching process. The etching process is highly anisotropic with quite smooth surface and vertical sidewalls.     

Effect of image resolution on intensity based scene illumination classification using neural network

In this paper, a framework for testing scene illumination classification with different image resolutions is proposed. The testing aims to provide the researchers with valuable information about the effect of image resolution on scene illumination classification using a neural network. The experiment is done by extracting three types of features from the images. These three types consist of statistical features, physic based features and histogram based features. It has been demonstrated that scene illumination classification can be affected by changing the image resolution. Despite the popular belief that high resolution images lead to better results, scene illumination classification by the proposed method performed best using low resolution images. At the second part of discussion, the reason behind this phenomenon is mathematically analysed and explained.     

The dry etching of a sol-gel deposited ZnO thin film in a high density BCl3/Ar plasma

The etching characteristics of zinc oxide (ZnO) including the etch rate and the selectivity of ZnO in a BCl₃/Ar plasma were investigated. It was found that the ZnO etch rate showed a non-monotonic behavior with an increasing BCl₃ fraction in the BCl₃/Ar plasma, along with the RF power, and gas pressure. At a BCl₃ (80%)/Ar (20%) gas mixture, the maximum ZnO etch rate of 50.3 nm/min and the maximum etch selectivity of 0.75 for ZnO/Si were obtained. Plasma diagnostics done with a quadrupole mass spectrometer delivered the data on the ionic species composition in plasma. Due to the relatively high volatility of the by-products formed during the etching by the BCl₃/Ar plasma, ion bombardment in addition to physical sputtering was required to obtain the high ZnO etch rates. The chemical state of the etched surfaces was investigated with X-ray Photoelectron Spectroscopy (XPS). Inferred from this data, it was suggested that the ZnO etch mechanism was due to ion enhanced chemical etching.     

Mechanical strength and setting times estimation of hydroxyapatite cement by using neural network

In this study, the mechanical strength, the initial and the final setting times in hydroxyapatite (HA) bone cement are estimated by designing a back-propagation neural network (BPNN) which has 2 inputs and 3 outputs. Firstly, some experimental samples have been prepared to train the BPNN to get it to estimate the output parameters. Then BPNN is tested using some experimental samples that have not been used in the training stage. To prepare the training and testing data sets, some experiments were performed. In these experiments, the β-tricalcium phosphate (β-TCP), the calcium carbonate and the dicalcium phosphate are used to prepare the powder part of the HA bone cement. Also the liquid part of the cement consists of the NaH₂PO₄⋅2H₂O solution with different concentrations. The effects of liquid phase concentration and the liquid/powder ratio of the cement, as input parameters, have been investigated on the setting times and the mechanical strength of the cement, as output parameters. The comparison of the predicted values and the experimental data indicates that the developed model has an acceptable performance to estimation of the setting times and the mechanical strength in HA bone cement. Also three neural networks with 2-inputs and 1-output was developed, similar to above method, and were compared with 3-outputs model. It is found that the prediction accuracy of 3-outputs model is better than those of other 1-output models.     

感应耦合等离子刻蚀InP工艺

采用Cl2/CH4/N2感应耦合等离子体对InP进行了刻蚀.系统地讨论了RF功率、ICP功率、反应腔压力、气体流量等工艺参数对InP材料端面刻蚀的影响.通过优化工艺参数,获得了光滑垂直的InP刻蚀端面,刻蚀速率达到841 nm/min,与SiO2的选择比达到15:1.     

Prediction of silicon dry etching using a piecewise linear algorithm

A piecewise linear algorithm for predicting silicon etch rates in fluorine-based plasmas is shown. Discrete experimental data of pressure and RF power in reactive ion etching are used to construct a set of local two-dimensional etching functions that serve as a basis for computing numerical solutions (pressure and power values for a specific predicted silicon etch rate). It must be pointed out that, although the algorithm scans the entire data domain, a testing procedure is applied to ensure that the computing task will be invoked only when a solution exists, and otherwise it will be discarded (this avoids brute force methods). In the last step of the algorithm, all solutions are collected and interpolated to construct a solution path. In order to verify the match between the experimental etching results and numerical predictions, the algorithm has been coded and tested using Maple® Release 13.0, showing a successful validation with a difference between experimental data and computed numerical solutions as low as 1% for SF₆, and 4% for SF₆/O₂ in the best case and a root-mean squared error of 0.03.