https://www2.econ.iastate.edu/tesfatsi/DeepLearningInNeuralNetworksOverview.JSchmidhuber2015.pdf

Deep learning in neural networks

Artificial Neural Networks (ANNs) are being used increasingly to predict and forecast water resources variables. In this paper, the steps that should be followed in the development of such models are outlined. These include the choice of performance criteria, the division and pre-processing of the available data, the determination of appropriate model inputs and network architecture, optimisation of the connection weights (training) and model validation. The options available to modellers at each of these steps are discussed and the issues that should be considered are highlighted. A review of 43 papers dealing with the use of neural network models for the prediction and forecasting of water resources variables is undertaken in terms of the modelling process adopted. In all but two of the papers reviewed, feedforward networks are used. The vast majority of these networks are trained using the backpropagation algorithm. Issues in relation to the optimal division of the available data, data pre-processing and the choice of appropriate model inputs are seldom considered. In addition, the process of choosing appropriate stopping criteria and optimising network geometry and internal network parameters is generally described poorly or carried out inadequately. All of the above factors can result in non-optimal model performance and an inability to draw meaningful comparisons between different models. Future research efforts should be directed towards the development of guidelines which assist with the development of ANN models and the choice of when ANNs should be used in preference to alternative approaches, the assessment of methods for extracting the knowledge that is contained in the connection weights of trained ANNs and the incorporation of uncertainty into ANN models.

Neural networks for the prediction and forecasting of water resources variables: a review of modelling issues and applications

This article is a brief review of dry etching as applied to pattern transfer, primarily in silicon technology. It focuses on concepts and topics for etching materials of interest in micromechanics. The basis of plasma-assisted etching, the main dry etching technique, is explained and plasma system configurations are described such as reactive ion etching (RIE). An important feature of RIE is its ability to achieve etch directionality. The mechanism behind this directionality and various plasma chemistries to fulfil this task will be explained. Multi-step plasma chemistries are found to be useful to etch, release and passivate micromechanical structures in one run successfully. Plasma etching is extremely sensitive to many variables, making etch results inconsistent and irreproducible. Therefore, important plasma parameters, mask materials and their influences will be treated. Moreover, RIE has its own specific problems, and solutions will be formulated. The result of an RIE process depends in a non-linear way on a great number of parameters. Therefore, a careful data acquisition is necessary. Also, plasma monitoring is needed for the determination of the etch end point for a given process. This review is ended with some promising current trends in plasma etching.

/pdf/a-survey-on-the-reactive-ion-etching-of-silicon-in-4a6u980s0l.pdf

A survey on the reactive ion etching of silicon in microtechnology

This paper discusses the principles and experimental status of gas cluster ion beam (GCIB) processing as a promising surface modification technique for practical industrial applications. Theoretical and experimental characteristics of GCIB processes and of related equipment development are described from the moment of neutral cluster formation, through ionization, acceleration and impact upon a surface. The impact of an accelerated cluster ion upon a target surface imparts very high energy densities into the impact area and produces non-linear effects that are not observed in the impacts of atomic ions. Unique characteristics of GCIB bombardment have been found to offer potential for various industrial applications that cannot be achieved by conventional ion beam processing. Among prospective applications are included shallow ion implantation, high rate sputtering, surface cleaning and smoothing, and low temperature thin film formation. Sputtering effects produced by cluster ion impact are particularly interesting. High sputtering yields and lateral distribution of sputtered atoms cause surface smoothing effects which cannot be achieved with monomer ion beams. Surface smoothing to atomic levels is expected to become the first production use of GCIB.

Materials Processing by Gas Cluster Ion Beams

Surface science aspects of etching reactions

The emission of halogen and alkali atoms, occurring under bombardment of alkali halides with electrons has been investigated. The electron energy was 540 eV and the temperature of the target was varied between room temperature and 400°C. We measured the energy distribution of the emitted neutral particles with a time of flight method. It was found that either diffusing interstitial halogen atoms or moving holes dominate the sputtering process above 200°C. Below 150°C alkali halides with lattice parameter s/d ≥ 0.33 show emission of non-thermal halogen atoms. s is the interionic space between two halogen ions in a direction and d is the diameter of a halogen atom. In general the energy distribution of the alkali and halogen atoms is thermal above 200°C, but not Maxwellian.

Energy distributions of atoms sputtered from alkali halides by 540 eV electrons

Argon‐ion assisted etching of silicon by molecular chlorine has been investigated. The masses and some kinetic energy distributions of the neutral particles emitted from the surface have been determined by mass spectrometry and time‐of‐flight spectra. It is found that an important part of the silicon‐containing particles consists of SiCl and SiCl2 molecules which have kinetic energies in the eV region. The observations exclude a simple evaporation of SiCl4 at the target temperature. A tentative model—consisting of a collision cascade like process parallel to thermal evaporation at the substrate temperature induced by a reduction of the effective surface binding energy during a sputtering event—is given to explain the observed kinetic energy distributions qualitatively.

Argon‐ion assisted etching of silicon by molecular chlorine

The sputtering of Si by 3‐keV Ar+ ions under simultaneous exposure to a beam of XeF2 gas has been investigated. The masses and some energy distributions of the neutral particles emitted from the target have been determined by mass spectrometry and time of flight. We found the following Si‐containing ions: Si+, SiF+, SiF+2, and SiF+3. The energy distributions of the sputtered particles cannot be explained by thermal desorption at the target temperature but can be interpreted as a collision cascadelike mechanism with a low surface binding energy.

Mass and energy distribution of particles sputter etched from Si in a XeF2 environment

Condensed gas layers of H2O, NH3 and CO at 15–20 K have been bombarded by 6 keV H+2 and 3 keV He+ and Ar+ ions. Mass spectra of the neutral species sputtered from these layers have been measured. There is a substantial yield of products which originally were not in the target material, and which have thus been formed in chemical reactions induced by the ion bombardment. The relative yields of some of the products have been found to increase with decreasing incident ion mass. This is mainly attributed to the larger amount of energy deposited by electronic stopping in such situations. From CO a nonvolatile residue is left after ion irradiation. From a layer of H2O frozen on top of the CO-residue H2CO was detected.

Reactive sputtering of simple condensed gases by keV ions II: Mass spectra

The reaction of Si with Cl2 alone and under simultaneous exposure to an Ar+ ion beam is investigated as a function of target temperature and Ar+ ion energy using mass spectrometry and time‐of‐flight studies. The main products of the thermal reaction are SiCl4 at low and SiCl2 at higher temperatures. The main products of combined exposure are atomic Si and Cl and molecular SiCl and SiCl2. The contributions of atoms and molecules are of a comparable order of magnitude. It appears that the main fraction of the products of the combined exposure results from sputtering and not from stimulation of the thermal reaction path by ion bombardment. The results strongly suggest that a modification of the top atomic layers of the Si, from which the products are sputtered, by incorporation of significant quantities of Cl plays a dominant part in the reaction mechanism.

A. Haring

Papers

Energy distributions of atoms sputtered from alkali halides by 540 eV electrons

Argon‐ion assisted etching of silicon by molecular chlorine

Mass and energy distribution of particles sputter etched from Si in a XeF2 environment

Reactive sputtering of simple condensed gases by keV ions II: Mass spectra

Ion‐assisted etching of silicon by molecular chlorine