Investigation of dead-layer thickness in SrRuO3/Ba0.5Sr0.5TiO3/Au thin-film capacitors

Thin-film capacitors, with barium strontium titanate (BST) dielectric layers between 7.5 and 950 nm in thickness, were fabricated by pulsed-laser deposition. Both crystallography and cation chemistry were consistent with successful growth of the BST perovskite. At room temperature, all capacitors displayed frequency dispersion such that epsilon (100 kHz)/epsilon (100 Hz) was greater than 0.75. The dielectric constant as a function of thickness was fitted, using the series capacitor model, for BST thicknesses greater than 70 nm. This yielded a large interfacial d(i)/epsilon (i) ratio of 0.40 +/-0.05 nm, implying a highly visible parasitic dead layer within the capacitor structure. Modeled consideration of the dielectric behavior for BST films, whose total thickness was below that of the dead layer, predicted anomalies in the plots of d/epsilon against d at the dead-layer thickness. In the capacitors studied here, no anomaly was observed. Hence, either (i) 7.5 nm is an upper limit for the total dead-layer thickness in the SRO/BST/Au system, or (ii) dielectric collapse is not associated with a distinct interfacial dead layer, and is instead due to a through-film effect. (C) 2001 American Institute of Physics.

A new Jacobian matrix for optimal learning of single-layer neural networks

This paper investigates the learning of a wide class of single-hidden-layer feedforward neural networks (SLFNs) with two sets of adjustable parameters, i.e., the nonlinear parameters in the hidden nodes and the linear output weights. The main objective is to both speed up the convergence of second-order learning algorithms such as Levenberg-Marquardt (LM), as well as to improve the network performance. This is achieved here by reducing the dimension of the solution space and by introducing a new Jacobian matrix. Unlike conventional supervised learning methods which optimize these two sets of parameters simultaneously, the linear output weights are first converted into dependent parameters, thereby removing the need for their explicit computation. Consequently, the neural network (NN) learning is performed over a solution space of reduced dimension. A new Jacobian matrix is then proposed for use with the popular second-order learning methods in order to achieve a more accurate approximation of the cost function. The efficacy of the proposed method is shown through an analysis of the computational complexity and by presenting simulation results from four different examples.

Functional characterization of the initiation enzyme of S-layer glycoprotein glycan biosynthesis in Geobacillus stearothermophilus NRS 2004/3a

The glycan chain of the S-layer glycoprotein of Geobacillus stearothermophilus NRS 2004/3a is composed of repeating units [-->2)-alpha-l-Rhap-(1-->3)-beta-l-Rhap-(1-->2)-alpha-l-Rhap-(1-->], with a 2-O-methyl modification of the terminal trisaccharide at the nonreducing end of the glycan chain, a core saccharide composed of two or three alpha-l-rhamnose residues, and a beta-d-galactose residue as a linker to the S-layer protein. In this study, we report the biochemical characterization of WsaP of the S-layer glycosylation gene cluster as a UDP-Gal:phosphoryl-polyprenol Gal-1-phosphate transferase that primes the S-layer glycoprotein glycan biosynthesis of Geobacillus stearothermophilus NRS 2004/3a. Our results demonstrate that the enzyme transfers in vitro a galactose-1-phosphate from UDP-galactose to endogenous phosphoryl-polyprenol and that the C-terminal half of WsaP carries the galactosyltransferase function, as already observed for the UDP-Gal:phosphoryl-polyprenol Gal-1-phosphate transferase WbaP from Salmonella enterica. To confirm the function of the enzyme, we show that WsaP is capable of reconstituting polysaccharide biosynthesis in WbaP-deficient strains of Escherichia coli and Salmonella enterica serovar Typhimurium.