No abstract

CdS was deposited onto clean cleaved InP(110) by molecular beam epitaxy (MBE) using a growth rate of 0.2 monolayers/min and a substrate temperature of 440 K (510 K). Raman spectra were taken in situ of the clean InP surface and after each evaporation step using an Ar+ ion laser as a light source. Due to this resonant excitation scattering signals originating from the CdS deposition are observed at coverages as low as 2 monolayers (ML). The number of phonon peaks observed and their selection rules reveal that the cubic modification is present. The spectra are dominated at all coverages by the longitudinal optical (LO) and 2LO phonon scattering intensities and the variation of the 2LO/LO intensity ratio with CdS deposition indicates changes in the electronic structure of the growing CdS. Another spectral feature in the Raman spectra is attributed to a chemically reacted layer at the interface most likely consisting of an In–S compound. The intensity of this feature is found to depend critically on the growth parameters, in particular the substrate temperature, but also on the operating time of the MBE cell. The amount of reaction at the interface also influences the critical CdS film thickness and the development of the 2LO/LO ratio. The results are discussed taking complementary photoluminescence, x-ray diffraction, and photoemission data into account.

A simple method for determining the hardness or thickness of very thin (few micrometre) ceramic coatings is described. The methods employed are demonstrated for hard, thin coatings of titanium nitride, deposited by physical vapour deposition on to high speed steel substrates. A procedure has been formulated, using the microhardness technique, to carry out thickness, hardness, and adhesion quality control checks on coated items.

No abstract

No abstract

No abstract

The sensitive information present in the training data, poses a privacy concern for applications as their unintended memorization during training can make models susceptible to membership inference and attribute inference attacks. In this paper, we investigate this problem in various pre-trained word embeddings (GloVe, ELMo and BERT) with the help of language models built on top of it. In particular, firstly sequences containing sensitive information like a single-word disease and 4-digit PIN are randomly inserted into the training data, then a language model is trained using word vectors as input features, and memorization is measured with a metric termed as exposure. The embedding dimension, the number of training epochs, and the length of the secret information were observed to affect memorization in pre-trained embeddings. Finally, to address the problem, differentially private language models were trained to reduce the exposure of sensitive information.

In this paper, we compare epitaxial Fe(Se,Te) thin films grown by pulsed laser deposition on CaF2, SrTiO3, MgO single crystals as well as on different metallic templates having a CeO2 based top surface. In particular, we performed a detailed structural and superconducting analysis. X-ray diffraction studies showed highly textured films on all templates. The superconducting transition temperatures are between 21 K and 14 K for Fe(Se,Te) films on CaF2 and on MgO single-crystal substrates, respectively, whereas films on the metal templates show T c values up to 18 K. The critical current density (J c) was determined from magnetization loops in fields up to 7 T. Calculations in the framework of the extended critical state model showed J c values over 2 MA cm−2 and 0.9 MA cm−2 at 5 K in self-field on single-crystal substrates and metal templates substrates, respectively. Detailed transmission electron microscopy studies reveal smooth layers on all templates. Whereas small defects were found in films on single crystals, pronounced grain boundaries with higher misorientation angle were visible in the layers on metallic templates.