Prediction of integral type failures in semiconductor manufacturing through classification methods

Smart management of maintenances has become fundamental in manufacturing environments in order to decrease downtime and costs associated with failures. Predictive Maintenance (PdM) systems based on Machine Learning (ML) techniques have the possibility with low added costs of drastically decrease failures-related expenses; given the increase of availability of data and capabilities of ML tools, PdM systems are becoming really popular, especially in semiconductor manufacturing. A PdM module based on Classification methods is presented here for the prediction of integral type faults that are related to machine usage and stress of equipment parts. The module has been applied to an important class of semiconductor processes, ion-implantation, for the prediction of ion-source tungsten filament breaks. The PdM has been tested on a real production dataset. © 2013 IEEE.

Nonreciprocal scattering by stacked nonlinear magneto-active semiconductor layers

The combinatorial frequency generation by the periodic stacks of magnetically biased semiconductor layers has been modelled in a self-consistent problem formulation, taking into account the nonlinear dynamics of carriers. It is shown that magnetic bias not only renders nonreciprocity of the three-wave mixing process but also significantly enhances the nonlinear interactions in the stacks, especially at the frequencies close to the intrinsic magneto-plasma resonances of the constituent layers. The main mechanisms and properties of the combinatorial frequency generation and emission from the stacks are illustrated by the simulation results, and the effects of the individual layer parameters and the structure arrangement on the stack nonlinear and nonreciprocal response are discussed. © 2014 Elsevier B.V. All rights reserved.

OPTIMIZATION OF THE SLOW-MODE PLASMON POLARITON IN LIGHT-EMITTING TUNNEL-JUNCTIONS

Light emitted from metal/oxide/metal tunnel junctions can originate from the slow-mode surface plasmon polariton supported in the oxide interface region. The effective radiative decay of this mode is constrained by competition with heavy intrinsic damping and by the need to scatter from very small scale surface roughness; the latter requirement arises from the mode's low phase velocity and the usual momentum conservation condition in the scattering process. Computational analysis of conventional devices shows that the desirable goals of decreased intrinsic damping and increased phase velocity are influenced, in order of priority, by the thickness and dielectric function of the oxide layer, the type of metal chosen for each conducting electrode, and temperature. Realizable devices supporting an optimized slow-mode plasmon polariton are suggested. Essentially these consist of thin metal electrodes separated by a dielectric layer which acts as a very thin (a few nm) electron tunneling barrier but a relatively thick (several 10's of nm) optically lossless region. (C) 1995 American Institute of Physics.

DEPENDENCE OF LIGHT OUTPUT FROM NOMINALLY SMOOTH AL-OX-AU TUNNEL-JUNCTIONS ON ELECTRODE MORPHOLOGY

The light output from nominally smooth Al-Ox-Au tunnel junctions is observed to be substantially independent of the deposition rate of the Au film electrode. Films deposited quickly (2 nm s-1) and those deposited slowly (0.16 nm s-1) have similar spectral dependences and intensities. (This is in contrast to roughened films where those deposited quickly give out less light, especially towards the blue end of the spectrum.) The behaviour can be interpreted in terms of the ratio l(ph)/l(em) where l(ph) and l(em) are the mean free paths of surface plasmons between external photon emissions and internal electromagnetic absorptions respectively. Once l(ph)/l(em) exceeds 100, as it does on smooth films, grain size has little further effect on the spectral shape of the light output. In fast-deposited films there are two compensating effects on the output intensity: grain boundary scattering decreases it and greater surface roughness increases it.

CHARACTERIZATION OF SURFACE-PLASMONS ON METAL-OXIDE-SEMICONDUCTOR STRUCTURES

Using the Otto (prism-air gap-sample) configuration p-polarized light of wavelength 632.8 nm has been coupled with greater than 80% efficiency to surface plasmons on the aluminium electrode of silicon-silicon dioxide-aluminium structures. The results show that if the average power per unit area dissipated on the metal film exceeds approximately 1 mW mm-2, then the coupling gap and thus the characteristics of the surface plasmon resonance are noticeably altered. In modelling the optical response of such systems the inclusion of both a non-uniform air coupling gap and a thin cermet layer at the aluminium surface may be necessary.

CHARACTERIZATION OF THE AU SURFACE-STRUCTURE OF AL-L-AU LIGHT-EMITTING TUNNEL-JUNCTIONS

Visible light is emitted from the Au-air interface of Al-I-Au thin-film tunnel junctions (deposited over a thin layer of CaF2 on glass) as a result of the decay of surface plasmon polaritons (SPPs). We show the surface topography of such a Au film and relate its large-scale features to the outcoupling of fast SPP's to photons. The absence of short-scale roughness features is explained by thier disappearance through surface diffusion. To confirm this a controlled sequence of 5-nm, 20-ms scanning tunneling microscope (STM) W tip crashes has been used to produce indentations 3 nm deep with a lateral dimension of 5-7 nm on a Au crystal in air at room temperature. Four sequences of indentations were drawn in the form of a square box. Right from the start, feature decay is observed and over a period of 2 h a succession of images shows that the structure disappears into the background as a result of surface diffusion. The surface diffusion constant is estimated to be 10(-18) cm2 s-1. The lack of light output via slow mode SPPs is an inevitable consequence of surface annealing.

Gaussian pulse mixing by stacked semiconductor layers

The nonlinear scattering of two Gaussian pulses with different central frequencies incident at slant angles on the periodic stack of binary semiconductor layers has been modelled in the self-consistent problem formulation taking into account the dynamics of charges. The effects of the pump pulse length and central frequencies, and the stack physical and geometrical parameters on the properties of the emitted combinatorial frequency waveforms are analysed and discussed.

Insight Extraction for Semiconductor Manufacturing Processes

In the semiconductor manufacturing environment it is very important to understand which factors have the most impact on process outcomes and to control them accordingly. This is usually achieved through design of experiments at process start-up and long term observation of production. As such it relies heavily on the expertise of the process engineer. In this work, we present an automatic approach to extracting useful insights about production processes and equipment based on state-of-the-art Machine Learning techniques. The main goal of this activity is to provide tools to process engineers to accelerate the learning-by-observation phase of process analysis. Using a Metal Deposition process as an example, we highlight various ways in which the extracted information can be employed.

On Regression Methods for Virtual Metrology in Semiconductor Manufacturing

Virtual metrology (VM) aims to predict metrology values using sensor data from production equipment and physical metrology values of preceding samples. VM is a promising technology for the semiconductor manufacturing industry as it can reduce the frequency of in-line metrology operations and provide supportive information for other operations such as fault detection, predictive maintenance and run-to-run control. The prediction models for VM can be from a large variety of linear and nonlinear regression methods and the selection of a proper regression method for a specific VM problem is not straightforward, especially when the candidate predictor set is of high dimension, correlated and noisy. Using process data from a benchmark semiconductor manufacturing process, this paper evaluates the performance of four typical regression methods for VM: multiple linear regression (MLR), least absolute shrinkage and selection operator (LASSO), neural networks (NN) and Gaussian process regression (GPR). It is observed that GPR performs the best among the four methods and that, remarkably, the performance of linear regression approaches that of GPR as the subset of selected input variables is increased. The observed competitiveness of high-dimensional linear regression models, which does not hold true in general, is explained in the context of extreme learning machines and functional link neural networks.

Inhibition of Rho-kinase protects cerebral barrier from ischaemia-evoked injury through modulations of endothelial cell oxidative stress and tight junctions

Ischaemic strokes evoke blood-brain barrier (BBB) disruption and oedema formation through a series of mechanisms involving Rho-kinase activation. Using an animal model of human focal cerebral ischaemia, this study assessed and confirmed the therapeutic potential of Rho-kinase inhibition during the acute phase of stroke by displaying significantly improved functional outcome and reduced cerebral lesion and oedema volumes in fasudil- versus vehicle-treated animals. Analyses of ipsilateral and contralateral brain samples obtained from mice treated with vehicle or fasudil at the onset of reperfusion plus 4 h post-ischaemia or 4 h post-ischaemia alone revealed these benefits to be independent of changes in the activity and expressions of oxidative stress- and tight junction-related parameters. However, closer scrutiny of the same parameters in brain microvascular endothelial cells subjected to oxygen-glucose deprivation ± reperfusion revealed marked increases in prooxidant NADPH oxidase enzyme activity, superoxide anion release and in expressions of antioxidant enzyme catalase and tight junction protein claudin-5. Cotreatment of cells with Y-27632 prevented all of these changes and protected in vitro barrier integrity and function. These findings suggest that inhibition of Rho-kinase after acute ischaemic attacks improves cerebral integrity and function through regulation of endothelial cell oxidative stress and reorganization of intercellular junctions. Inhibition of Rho-kinase (ROCK) activity in a mouse model of human ischaemic stroke significantly improved functional outcome while reducing cerebral lesion and oedema volumes compared to vehicle-treated counterparts. Studies conducted with brain microvascular endothelial cells exposed to OGD ± R in the presence of Y-27632 revealed restoration of intercellular junctions and suppression of prooxidant NADPH oxidase activity as important factors in ROCK inhibition-mediated BBB protection.

Pulse mixing by Fibonacci and Thue-Morse ordered semiconductor layers

The properties of combinatorial frequency generation by two-tone Gaussian pulses incident at oblique angles on quasiperiodic (Fibonacci and Thue-Morse) stacks of binary semiconductor layers are discussed. The analysis has been performed using the self-consistent model taking into account the nonlinear dynamics of mobile charges in the layers. The effects of the stack arrangements and constituent layer parameters on the combinatorial frequency waveforms are presented for the specific structures of both types