Modeling and optimization of semiconductor manufacturing process with neural networks

A neural network-based process model is proposed to optimize the semiconductor manufacturing process. Being different from some works in several research groups which developed neural network-based models to predict process quality with a set of process variables of only single manufacturing step, we applied this model to wafer fabrication parameters control and wafer lot yield optimization. The original data are collected from a wafer fabrication line, including technological parameters and wafer test results. The wafer lot yield is taken as the optimization target. Learning from historical technological records and wafer test results, the model can predict the wafer yield. To eliminate the "bad" or noisy samples from the sample set, an experimental method was used to determine the number of hidden units so that both good learning ability and prediction capability can be obtained.

Ultrafast low-temperature grown AlGaAs/GaAs photorefractive quantum wells using point defects as capture centers

At a medium substrate temperature of 400 degrees C and a lower As flux, we have grown an ultrafast AlGaAs/GaAs photorefractive multiple quantum well (MQW) structure by molecular beam epitaxy. The as-grown sample exhibits strong photorefractive effect under the transverse Frantz-Keldysh geometry. A peak electroabsorption of 2100 cm(-1) is measured in the as-grown sample in an 11 kV/cm dc electric field, and the peak photorefractive diffraction efficiency can be 1.2%. After postgrowth annealing, the photorefractive effect becomes weak and disappears in samples annealed above 700 degrees C. Using optical transient current spectroscopy, deep levels are measured in these samples. It is found that deep levels are stable against annealing until 700 degrees C. Using a pump-probe technique, carrier lifetimes are measured at room temperature. We find that the as-grown sample has a lifetime of 20 ps, while the 700 degrees C annealed sample has a lifetime of more than 200 ps. The ultrafast lifetime in the as-grown sample is caused by point defects, not by As clusters. Our result show that AlGaAs/GaAs MQW structure grown around 400 degrees C has better performance of the photorefractive effect. (C) 1999 American Institute of Physics. [S0003-6951(99)04036-X].

Effect of defects on optical phonon Raman spectra in SiC nanorods

Fabricated one-dimensional (1D) materials often have abundant structural defects. Experimental observation and numerical calculation indicate that the broken translation symmetry due to structural defects may play a more important role than the quantum confinement effect in the Raman features of optical phonons in polar semiconductor quantum wires such as SiC nanorods, (C) 1999 Elsevier Science Ltd. All rights reserved.

Influence of (001) vicinal GaAs substrates on the optical properties of defects in low-temperature grown GaAs/AlGaAs multiple quantum wells

Photoluminescence (PL) spectroscopy and carrier lifetime measurement has been used to characterize optical properties of defects in the low-temperature (LT) grown GaAs/AlGaAs multiple quantum well structures. Two sets of samples were grown at 400 degrees C by molecular beam epitaxy on nominal (001) and miscut [4 degrees off (001) towards (111) A] GaAs substrates, respectively. After growth, samples were subjected to 30 s rapid thermal annealing at 600-800 degrees C. It is found that after annealing, two defect-related PL features appear in the samples grown on nominal (001) GaAs substrates, but not in those grown on miscut (001) GaAs substrates. The carrier lifetimes are about 31 and 5 ps in as-grown samples grown on nominal and miscut (001) GaAs substrates, respectively. The different PL spectra and carrier lifetimes in two sets of samples are attributed to different structures of the As-Ga-like defects formed during LT growth. (C) 1999 American Institute of Physics. [S0003-6951(99)00230-2].

Void-like defects in annealed Czochralski silicon

Void-like defects of octahedron structure having {111} facets were observed in annealed Czochralski silicon. The amorphous coverage of SiOx and SiCx on the inner surface of the defects was identified using transmission electron microscopy and electron energy-loss spectroscopy. It is suggested that these defects are a kind of amorphous precipitate origin. A mechanism for the generation of these defects and the previously reported solid amorphous precipitates is proposed. (C) 1998 American Institute of Physics. [S0003-6951(98)02842-3].

Observation of defects in GaN epilayers

GaN epilayers grown on sapphire substrates nitridated for various lengthy periods were investigated by light scattering tomography (LST) and Raman scattering. In the LST images of the plane-view epilayers, the light scattering defects distribute in [<11(2)over bar 0>] directions. The defect density is lower in epilayer grown on substrate nitridated for a longer period. The defects are believed to be straight threading edge dislocations on {<1(1)over bar 00>} planes. The Raman shift of E-2 mode is larger in the sample grown on substrate nitridated for a longer period. Our results show that the stress is higher in the sample with fewer dislocations.

Hydrogen-decorated lattice defects in proton implanted GaN

Several vibrational bands were observed near 3100 cm(-1) in GaN that had been implanted with hydrogen at room temperature and subsequently annealed, Our results indicate that these bands are due to nitrogen-dangling-bond defects created by the implantation that an decorated by hydrogen, The frequencies are close to those predicted recently for V-Ga-H-n complexes, leading us to tentatively assign the new lines to V-Ga defects decorated with different numbers of H atoms. (C) 1998 American Institute of Physics. [S0003-6951(98)03614-6].

Native deep level defects in ZnO single crystal grown by CVT method - art. no. 68410I

Hall effect, photoluminescence (PL), infrared absorption, deep level transient spectroscopy (DLTS), and Raman scattering have been used to study property and defects of ZnO single crystal grown by a chemical vapor transport method (CVT). As-grown ZnO is N type with free electron density Of 10(16)-10(17)cm(-3). It has a slight increase after 900 degrees C annealing in oxygen ambient. The DLTS measurement revealed four deep level defects with energy at 0.30eV, 0.50eV, 0.68eV and 0.90eV in the as-grown ZnO sample, respectively. After the high temperature annealing, only the 0.5eV defect survive and has a concentration increase. PL results of the as-grown and annealed ZnO indicate that the well-known green emission disappear after the annealing. The result suggests a correlation between the 0.68eV defect and the green PL peak. Results of P-doped ZnO were also compared with the undoped ZnO sample. The nature of the defects and their influence on the material property have been discussed.

Structure approximate based on double weighted neural network

Double weighted neural network; is a kind of new general used neural network, which, compared with BP and RBF network, may approximate the training samples with a move complicated geometric figure and possesses a even greater approximation. capability. we study structure approximate based on double weighted neural network and prove its rationality.

Dynamic power management approaches based on neural network

Dynamic Power Management (DPM) is a technique to reduce power consumption of electronic system. by selectively shutting down idle components. In this article we try to introduce back propagation network and radial basis network into the research of the system-level policies. We proposed two PAY policies-Back propagation Power Management (BPPM) and Radial Basis Function Power management (RBFPM) which are based on Artificial Neural Networks (ANN). Our experiments show that the two power management policies greatly lowered the system-level power consumption and have higher performance than traditional Power Management(PM) techniques-BPPM is 1.09-competitive and RBFPM is 1.08-competitive vs. 1.79,145,1.18-competitive separately for traditional timeout PM, adaptive predictive PM and stochastic PM.

Cancer classification based on direction-basis-function neural networks

Automatic molecular classification of cancer based on DNA microarray has many advantages over conventional classification based on morphological appearance of the tumor. Using artificial neural networks is a general approach for automatic classification. In this paper, Direction-Basis-Function neuron and Priority-Ordered algorithm are applied to neural networks. And the leukemia gene expression dataset is used as an example to testify the classifier. The result of our method is compared to that of SVM. It shows that our method makes a better performance than SVM.

Low Power Integrated Circuits for Wireless Neural Recording Applications

A group of prototype integrated circuits are presented for a wireless neural recording micro-system. An inductive link was built for transcutaneous wireless power transfer and data transmission. Power and data were transmitted by a pair of coils on a same carrier frequency. The integrated receiver circuitry was composed of a full-wave bridge rectifier, a voltage regulator, a date recovery circuit, a clock recovery circuit and a power detector. The amplifiers were designed with a limited bandwidth for neural signals acquisition. An integrated FM transmitter was used to transmit the extracted neural signals to external equipments. 16.5 mW power and 50 bps - 2.5 Kbps command data can be received over 1 MHz carrier within 10 mm. The total gain of 60 dB was obtained by the preamplifier and a main amplifier at 0.95Hz - 13.41 KHz with 0.215 mW power dissipation. The power consumption of the 100 MHz ASK transmitter is 0.374 mW. All the integrated circuits operated under a 3.3 V power supply except the voltage regulator.

Multi channel micro neural probe fabricated with SOI

Extracellular neural recording requires neural probes having more recording sites as well as limited volumes. With its mechanical characteristic and abundant process method, Silicon is a kind of material fit for producing neural probe. Silicon on insulator (SOI) is adopted in this paper to fabricate neural probes. The uniformity and manufacturability are improved. The fabricating process and testing results of a series of Multi channel micro neural probes were reported. The thickness of the probe is 15 mu m-30 mu m. The typical impedance characteristics of the record sites are around 2M Omega at 1k Hz. The performance of the neural probe in-vivo was tested on anesthetic rat. The recorded neural spike was typically around 140 mu V. Spike recorded from individual site could exceed 700 mu V. The average signal noise ratio was 7 or more.

Implantable Microsystem for Wireless Neural Recording Applications

A prototype microsystem is presented for wireless neural recording application. An inductive link was built for transcutaneous wireless power transfer and data transmission. Total 16.5 mW power and 50 bps - 2.5 Kbps command data can be received over 1 - 5 MHz with a distance of 0-10 mm. The integrated amplifiers were designed with a limited bandwidth for neural signals acquisition. The gain of 60 dB was obtained by preamplifier at 7 Hz - 3 KHz. An integrated FM transmitter was used to transmit the extracted neural signals to external equipments with 0.374 - 2 mW power comsumption and a maximum data rate of 500 Kbps at 100 MHz. All the integrated circuits modules except the power recovery circuit were tested or stimulated under a 3.3 V power supply, and fabricated in standard CMOS processing.

SOC dynamic power management using artificial neural network

Dynamic Power Management (DPM) is a technique to reduce power consumption of electronic system by selectively shutting down idle components. In this article we try to introduce back propagation network and radial basis network into the research of the system-level power management policies. We proposed two PM policies-Back propagation Power Management (BPPM) and Radial Basis Function Power Management (RBFPM) which are based on Artificial Neural Networks (ANN). Our experiments show that the two power management policies greatly lowered the system-level power consumption and have higher performance than traditional Power Management(PM) techniques-BPPM is 1.09-competitive and RBFPM is 1.08-competitive vs. 1.79 . 1.45 . 1.18-competitive separately for traditional timeout PM . adaptive predictive PM and stochastic PM.