Modelling of plasma etching using a generalized regression neural network

Plasma etching was modelled by using a generalized regression neural network (GRNN). The etching process was characterized with a statistical experimental design. Three etch responses were modelled, which include two etch rates of aluminium and silica and etching profile. GRNN prediction ability was optimized as a function of training factor. Three types of models were constructed depending on the type of prepared data. Type I model corresponds to the model constructed with the original, non-classified data. Type II and III models were built for the classified data without and with the control of data interface, respectively. Compared to type I models, type II models for two etch rates demonstrated more than 25% improvement. By the control of data interface, type III models exhibited more than 15% improvement over type II models. Classification-based models in conjunction with data control thus illustrated much improved prediction of GRNN over those for non-classified models.     

On the use of a neural network to characterize the plasma etching of SiON thin films

Using a generalized regression neural network (GRNN), plasma etching of oxynitride thin films was modeled. The etch process was characterized by means of a statistical experiment. A genetic algorithm was employed to improve prediction performance by optimizing multiparameterized training factors. Compared to a conventional GRNN model, the constructed etch rate model demonstrated an improvement of about 60% in the prediction performance. 3-D plots were generated to qualitatively interpret etch mechanisms while validating the predictions with experimental data. In separating physical and chemical effects, both dc bias and profile angle variations were effectively utilized. The source power affected significantly the etch rate irrespective of changes in the bias power or C₂F₆ flow rate. For pressure variations, the etch rate was estimated to be dominated by chemical etching. The complex effect of C₂F₆ flow rate could be explained by dominant chemical etching or polymer deposition.     

Optimization of principal-component-analysis-applied in situ spectroscopy data using neural networks and genetic algorithms

A new model of multidimensional in situ diagnostic data is presented. This was accomplished by combining a back-propagation neural network (BPNN), principal component analysis (PCA), and a genetic algorithm (GA). The PCA was used to reduce input dimensionality. The GA was applied to search for a set of optimized training factors involved in BPNN training. The presented technique was evaluated with optical emission spectroscopy (OES) data measured during the etching of oxide thin films in a CHF(3)-CF(4) inductively coupled plasma. For a systematic modeling, the etching process was characterized by a face-centered Box Wilson experiment. The etch responses to be modeled include oxide etch rate, oxide profile angle, and oxide etch rate non-uniformity. In PCA, three types of data variances were employed and the reduced input dimensionality corresponding to 100, 99, and 98% are 16, 8, and 5. The BPNN training factors to be optimized include the training tolerance, number of hidden neurons, magnitude of initial weight distribution, gradient of bipolar sigmoid function, and gradient of linear function. The prediction errors of GA-BPNN models are 249 A/min, 2.64 degrees, and 0.439% for the etch rate, profile angle, and etch rate non-uniformity, respectively. Compared to the conventional and previous full OES models, the presented models demonstrated a significantly improved prediction for all etch responses.     

Development and analyses of data-driven models for predicting the bed depth profile of solids flowing in a rotary kiln

Soft computing data-driven modeling (DDM) techniques have attracted the attention of many researchers across the globe as they do not require deep knowledge of the complex physical process. In the present research, data-driven based models have been developed using support vector regression (SVR), multilayer perceptron neural network (MLP), radial basis function neural network (RBFNN) and general regression neural networks (GRNN) techniques for predicting the bed depth profile of solids flowing in a rotary kiln. The performances of the developed models were compared and evaluated against the experimental results in terms of statistical measures such as coefficient of determination (R²), and average absolute relative error (AARE). The obtained results and findings from this research have revealed that data-driven models can predict the bed depth profile of solids flowing in a rotary kiln quite accurately. The SVR-based model exhibited the lowest AARE value of 1.72% and highest R² value of 0.9981 while GRNN, RBFNN, and MLP models gave corresponding values of AARE as 3.69%, 55.13%, 98.15% and those of R² as 0.9898, 0.0052 and 0.0081, respectively. Moreover, the developed DDM-based models i.e. GRNN, RBFNN, and MLP models overcame the limitations of the existing solutions which involved iterative numerical procedure entailing high degree of computational complexity.     

From laboratory- to pilot-scale: moisture monitoring in fluidized bed granulation by a novel microwave sensor using multivariate calibration approaches

Recently, microwave resonance technology (MRT) sensor systems operating at four resonances instead of a single resonance frequency were established as a process analytical technology (PAT) tool for moisture monitoring. The additional resonance frequencies extend the technologies’ possible application range in pharmaceutical production processes remarkably towards higher moisture contents. In the present study, a novel multi-resonance MRT sensor was installed in a bottom-tangential-spray fluidized bed granulator in order to provide a proof-of-concept of the recently introduced technology in industrial pilot-scale equipment. The mounting position within the granulator was optimized to allow faster measurements and thereby even tighter process control. As the amount of data provided by using novel MRT sensor systems has increased manifold by the additional resonance frequencies and the accelerated measurement rate, it permitted to investigate the benefit of more sophisticated evaluation methods instead of the simple linear regression which is used in established single-resonance systems. Therefore, models for moisture prediction based on multiple linear regression (MLR), principal component regression (PCR), and partial least squares regression (PLS) were built and assessed. Correlation was strong (all R²?>?0.988) and predictive abilities were rather acceptable (all RMSE ≤0.5%) for all models over the whole granulation process up to 16% residual moisture. While PCR provided best predictive abilities, MLR proofed as a simple and valuable alternative without the need of chemometric data evaluation.     

Model-based analysis of the silica glass film etch mechanism in CF4/O2 inductively coupled plasma

The analysis of the Er-doped silica glass films (62%SiO₂–30%B₂O₃–8%P₂O₅ + 0.2 wt%. Er₂O₃) etch mechanism in the CF₄/O₂ inductively coupled plasma was carried out using the combination of simplified models for plasma chemistry and etch kinetics. As the O₂ mixing ratio in the CF₄/O₂ plasma increases from 0% to 30%, the etch rate decreases monotonically in the range of 385–190 nm/min that contradicts with the behavior of F atom density and flux. From the model-based analysis, it was found that, at low ion bombardment energies, the etch process followed the formal kinetics of ion-assisted chemical reaction and was controlled by both neutral and ion fluxes.     

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.     

Planar plasma etching of chromium

We have investigated the plasma etching characteristics of chromium thin films in an rf planar (parallel plate) reactor. The experimental work was performed using a commercial reactor operating at 13.56 MHz with power variable up to 500 W. The etch rate of the Cr films deposited on glass substrates by e-beam evaporation was measured as a function of the concentration of O₂ in a CCl₄/O₂ gas mixture, the total flow rate of input gases and the rf power density. Using a total gas flow of 15 sccm and an input power density of 0.4 W cm^?2, the maximum etch rate was obtained in CCl₄ plasma containing 40% O₂. It was found that doubling the number of the substrates in the reactor decreased the etch rate by 20%. Also, the etch rate at the back of the reactor was twice that at the front. Methods to alliviate non-uniformity and loading effects are discussed and the mechanism of plasma etching of Cr is examined through the effect of various processing parameters on the etching characteristics.     

Prediction of surface roughness using X-ray photoelectron spectroscopy and neural networks

A new model for controlling plasma processes was constructed by combining atomic force microscopy (AFM), X-ray photoelectron spectroscopy (XPS), and neural networks. The applicability of XPS to modeling etch rate was also investigated, as well as the impact of dc bias inclusion. The back-propagation neural network was used to find complex relationships between XPS and AFM data. This technique was evaluated with the etching data characterized by a 2(4) full factorial experiment. Five prediction models of surface roughness were constructed and compared. The Type I model refers to the model constructed with conventional process parameters. The Type II and III models were built with XPS and XPS plus dc bias data, respectively. The remaining Type IV and V models refer to those constructed with principal component analysis (PCA) reduced-XPS and PCA reduced-XPS plus dc bias, respectively. Mode prediction performance was evaluated as a function of training factor. In predicting the surface roughness, the Type II model yielded an improved prediction of 39% with respect to the Type IV model. The improvement was also demonstrated in modeling the etch rate. These results indicate that utilizing full XPS data is more effective for improving the model prediction performance. The advantage of XPS data was more conspicuous in constructing the surface roughness model.     

Etch characteristics of HfO2 thin films by using CF4/Ar inductively coupled plasma

In this study, we carried out an investigation of the etching characteristics (etch rate, selectivity) of HfO₂ thin films in the CF₄/Ar inductively coupled plasma (ICP). The maximum etch rate of 54.48 nm/min for HfO₂ thin films was obtained at CF₄/Ar (=20:80%) gas mixing ratio. At the same time, the etch rate was measured as function of the etching parameters such as ICP RF power, DC-bias voltage, and process pressure. The X-ray photoelectron spectroscopy analysis showed an efficient destruction of the oxide bonds by the ion bombardment as well as an accumulation of low volatile reaction products on the etched surface. Based on these data, the chemical reaction was proposed as the main etch mechanism for the CF₄-containing plasmas.     

Wavelet neural networks for modelling high pressure inactivation kinetics of Listeria monocytogenes in UHT whole milk

The aim of the present work was to investigate the applicability of a Wavelet Neural Network to describe the inactivation pattern of Listeria monocytogenes by high hydrostatic pressure in ultra high temperature (UHT) whole milk, and evaluate its performance against models used in predictive microbiology such as the re-parameterized Gompertz and modified Weibull equations. A comparative study with linear partial least squares regression (PLS-R) as well as neural network (NN) models demonstrated on the same dataset has been also considered. Milk was artificially inoculated with an initial population of the pathogen of ca. 10⁷ CFU/ml and exposed to a range of high pressures (350, 450, 550, 600 MPa) for up to 40 min at ambient temperature (ca. 25 °C). Typical survival curves were obtained including a shoulder, a log-linear and a tailing phase. Increasing the magnitude of the applied pressure resulted in increasing levels of inactivation. Modelling approaches provided good fit to experimental training data as inferred by the low values of the root mean squared error (RMSE) and the high values of regression coefficient (R²). Models were validated at 400 and 500 MPa with independent experimental data. First or second order polynomial models were employed to relate the inactivation parameters to pressure, whereas the wavelet network as well as the PLS and NN models were utilised as a one-step modelling approach. The prediction performance of the proposed learning-based network was better at both validation pressures. The development of accurate models to describe the survival curves of micro-organisms in high pressure treatment would be very important to the food industry for process optimisation, food safety and would eventually expand the applicability of this non-thermal process.     

Hydrogen plasma etching of silicon dioxide in a hollow cathode system

Chemical etching of various materials has been observed when hydrogen plasmas are used in material processing. In the case of the deposition of diamond films the preferential etching of sp² bonded carbon is considered to be of fundamental importance. A few papers have been published which have indicated that etching by hydrogen ions is different to that by hydrogen atoms. In this paper we describe the etching of silicon dioxide by hydrogen which was plasma-activated in a molybdenum-lined RF hollow cathode. The etch rate was seen to be thermally activated but decreased with increasing plasma power. The addition of a few percentage of helium increased the etch rate. The application of a − 50 V bias to the sample holder almost doubled the etch rate indicating the importance of ion bombardment for the chemical reaction. At high plasma powers and substrate temperatures in excess of 450 °C a thin molybdenum deposit was formed on the quartz samples.     

基于GRNN的金山店铁矿爆破振动峰值速度预测

为研究爆破振动对金山店铁矿地表构筑物和井下巷道的影响,引入广义回归神经网络(GRNN)的方法,分别以地表、井下部分振动监测数据为学习样本对GRNN进行训练,构建地表、井下爆破振动峰值速度的GRNN预测模型,以剩余振动监测数据为检测样本对地表和井下GRNN预测模型进行检验,并将GRNN模型的预测结果与BPNN、基函数回归法和经验公式法作对比。同时,将地表、井下GRNN模型的预测结果与以地表和井下综合训练数据为学习样本构建的综合GRNN预测模型进行对比。研究结果表明:对于地表监测点,四种方法的预测误差率依次为12.1%、18.9%、30.3%、43.7%;对于井下监测点,四种方法的预测误差率依次为14.0%、16.2%、19.9%、23.0%。GRNN的预测精度最高,其为爆破振动峰值速度的预测提供了一种新方法,且采用GRNN对地表、井下质点爆破振动峰值速度分别进行预测更加合理。     

Dry etching of TiN in N2/Cl2/Ar adaptively coupled plasma

We investigated the N₂ additive effect on the etch rates of TiN and SiO₂ and etch profile of TiN in N₂/Cl₂/Ar adaptively coupled plasma (ACP). The mixing ratio of Cl₂ and Ar was fixed at 75 and 25 sccm, respectively. The N₂ flow rate was increased from 0 to 9 sccm under the constant pressure of 10 mTorr. As N₂ flow rate was increased in N₂/Cl₂/Ar plasma, the etch rate of TiN was linearly increased, but that of SiO₂ was increased non-monotonically. The etch profile and the compositional changes of TiN was investigated with field emission-scanning electron microscope (FE-SEM), FE-Auger electron spectroscopy (FE-AES) and x-ray photoelectron spectroscopy (XPS). When 9 sccm N₂ was added into Cl₂/Ar, a steep etch profile and clean surface of TiN was obtained. In addition, the signals of TiN and Ti were disappeared in FE-AES and XPS when N₂ additive flow into Cl₂/Ar was above 6 sccm. From the experimental data, the increase in TiN etch rate was mainly caused by the increase of desorption and evacuation rate of etch by products because of the increased effective pumping speed. The etch mechanism of TiN in N₂/Cl₂/Ar ACP plasma can be concluded as the ion enhanced chemical etch.     

Dry etching of TiN in N2/Cl2/Ar adaptively coupled plasma

We investigated the N₂ additive effect on the etch rates of TiN and SiO₂ and etch profile of TiN in N₂/Cl₂/Ar adaptively coupled plasma (ACP). The mixing ratio of Cl₂ and Ar was fixed at 75 and 25 sccm, respectively. The N₂ flow rate was increased from 0 to 9 sccm under the constant pressure of 10 mTorr. As N₂ flow rate was increased in N₂/Cl₂/Ar plasma, the etch rate of TiN was linearly increased, but that of SiO₂ was increased non-monotonically. The etch profile and the compositional changes of TiN was investigated with field emission-scanning electron microscope (FE-SEM), FE-Auger electron spectroscopy (FE-AES) and x-ray photoelectron spectroscopy (XPS). When 9 sccm N₂ was added into Cl₂/Ar, a steep etch profile and clean surface of TiN was obtained. In addition, the signals of TiN and Ti were disappeared in FE-AES and XPS when N₂ additive flow into Cl₂/Ar was above 6 sccm. From the experimental data, the increase in TiN etch rate was mainly caused by the increase of desorption and evacuation rate of etch by products because of the increased effective pumping speed. The etch mechanism of TiN in N₂/Cl₂/Ar ACP plasma can be concluded as the ion enhanced chemical etch.     

Joint multiple adaptive wavelet regression ensembles

Multiple adaptive discrete wavelet transforms were applied during a multiple regression of spectroscopic data for the purpose of investigating the hypothesis — does the use of different wavelets, at different points, within a spectrum, elucidate predictive capability. The model investigated was a constrained stacking regression ensemble with individual regression models chosen initially by a Bayes Metropolis search. The ensemble approach provided the ability to combine different regression models that used different types of wavelets. Models were applied to a publically available dataset, pertaining to biscuit dough, of near infrared spectra, that were measured by a FOSS 5000, and laboratory measurements of the fat, flour, sugar and moisture content.The resultant model, which is referred to as a joint multiple adaptive wavelet regression ensemble (JMAWRE), was found to be the superior predictive model when compared to models that used standard wavelets as part of the regression ensembles. The JMAWRE was also superior when compared to other models from literature that used the same publicly available NIR dataset.     

Nanotexturing of GaN light-emitting diode material through mask-less dry etching

We describe a new technique for random surface texturing of a gallium nitride (GaN) light-emitting diode wafer through a mask-less dry etch process. This involves depositing a sub-monolayer film of silica nanospheres (typical diameter of 200 nm) and then subjecting the coated wafer to a dry etch process with enhanced physical bombardment. The silica spheres acting as nanotargets get sputtered and silica fragments are randomly deposited on the GaN epi-layer. Subsequently, the reactive component of the dry etch plasma etches through the exposed GaN surface. Silica fragments act as nanoparticles, locally masking the underlying GaN. The etch rate is much reduced at these sites and consequently a rough topography develops. Scanning electron microscopy (SEM) and atomic force microscopy (AFM) inspections show that random topographic features at the scale of a few tens of nanometres are formed. Optical measurements using angle-resolved photoluminescence show that GaN light-emitting diode material thus roughened has the capability to extract more light from within the epilayers.     

Utilizing support vector machine in real-time crash risk evaluation

Real-time crash risk evaluation models will likely play a key role in Active Traffic Management (ATM). Models have been developed to predict crash occurrence in order to proactively improve traffic safety. Previous real-time crash risk evaluation studies mainly employed logistic regression and neural network models which have a linear functional form and over-fitting drawbacks, respectively. Moreover, these studies mostly focused on estimating the models but barely investigated the models’ predictive abilities. In this study, support vector machine (SVM), a recently proposed statistical learning model was introduced to evaluate real-time crash risk. The data has been split into a training dataset (used for developing the models) and scoring datasets (meant for assessing the models’ predictive power). Classification and regression tree (CART) model has been developed to select the most important explanatory variables and based on the results, three candidates Bayesian logistic regression models have been estimated with accounting for different levels unobserved heterogeneity. Then SVM models with different kernel functions have been developed and compared to the Bayesian logistic regression model. Model comparisons based on areas under the ROC curve (AUC) demonstrated that the SVM model with Radial-basis kernel function outperformed the others. Moreover, several extension analyses have been conducted to evaluate the effect of sample size on SVM models’ predictive capability; the importance of variable selection before developing SVM models; and the effect of the explanatory variables in the SVM models. Results indicate that (1) smaller sample size would enhance the SVM model's classification accuracy, (2) variable selection procedure is needed prior to the SVM model estimation, and (3) explanatory variables have identical effects on crash occurrence for the SVM models and logistic regression models.     

The dry etching property of TiO2 thin films using metal-insulator-metal capacitor in inductively coupled plasma system

In this work, we investigated the etching characteristics of TiO₂ thin films and the selectivity of TiO₂ to SiO₂ in a BCl₃/Ar inductively coupled plasma (ICP) system. The maximum etch rate of 84.68 nm/min was obtained for TiO₂ thin films at a gas mixture ratio of BCl₃/Ar (25:75%). In addition, etch rates were measured as a function of etching parameters, such as the RF power, DC-bias voltage and process pressure. Using the X-ray photoelectron spectroscopy analysis the accumulation of chemical reaction on the etched surface was investigated. Based on these data, the ion-assisted physical sputtering was proposed as the main etch mechanism for the BCl₃-containing plasmas.