Recent advances in Large Language Model (LLM) based Generative AI techniques have made it feasible to translate enterprise-level code from legacy languages such as COBOL to modern languages such as Java or Python. While the results of LLM-based automatic transformation are encouraging, the resulting code cannot be trusted to correctly translate the original code, making manual validation of translated Java code from COBOL a necessary but time-consuming and labor-intensive process. In this paper, we share our experience of developing a testing framework for IBM Watsonx Code Assistant for Z (WCA4Z) [5], an industrial tool designed for COBOL to Java translation. The framework automates the process of testing the functional equivalence of the translated Java code against the original COBOL programs in an industry context. Our framework uses symbolic execution to generate unit tests for COBOL, mocking external calls and transforming them into JUnit tests to validate semantic equivalence with translated Java. The results not only help identify and repair any detected discrepancies but also provide feedback to improve the AI model.

Recent advances in Large Language Model (LLM) based Generative AI techniques have made it feasible to translate enterprise-level code from legacy languages such as COBOL to modern languages such as Java or Python. While the results of LLM-based automatic transformation are encouraging, the resulting code cannot be trusted to correctly translate the original code. We propose a framework and a tool to help validate the equivalence of COBOL and translated Java. The results can also help repair the code if there are some issues and provide feedback to the AI model to improve. We have developed a symbolic-execution-based test generation to automatically generate unit tests for the source COBOL programs which also mocks the external resource calls. We generate equivalent JUnit test cases with equivalent mocking as COBOL and run them to check semantic equivalence between original and translated programs. Demo Video: https://youtu.be/aqF_agNP-lU

Feature profiles of Si etched in HBr-containing plasmas have been analyzed through a comparison between experiments and simulations. The emphasis was placed on a mechanistic understanding of the difference in the evolution of profile anomalies (such as tapering, footing, and microtrenching) during Si etching between HBr- and Cl-2-based plasmas. Experiments were made with Cl-2/O-2/HBr chemistry by varying the HBr mixing ratio, using a commercial ultrahigh-frequency electron cyclotron resonance plasma etching reactor, where HCl/O-2 chemistry was also employed to compare with that of Cl-2/O-2 and HBr/O-2. Numerical simulations of feature profile evolution were made using a semiempirical atomic-scale cellular model based on the Monte Carlo method that we developed for Si etching in Br-2, HBr, and Cl-2 plasmas, where surface chemistry and kinetics include the effects of ion reflection from and/or penetration into feature surfaces on incidence. The experiments showed more vertical sidewalls with less footing and microtrenching with HBr; concretely, with increasing HBr mixing ratio in Cl-2/O-2/HBr plasmas, the tapering is reduced and minimized at 80% HBr where slight lateral or side etching tends to occur, the footing is reduced gradually, and the microtrenching fades away at more than 20% HBr. A comparison with simulations, with the help of separate analyses of ion reflection from surfaces on incidence, indicated that the smaller reflection probability and reflected energy fraction of Br+ on tapered sidewalls (compared to Cl+) are responsible for reduced tapering, footing, and microtrenching in HBr-containing plasmas; moreover, chemical etching effects of neutral H atoms at the feature bottom and sidewalls, arising from the larger reaction probability of H (compared to Cl), are also responsible for reduced microtrenching and for reduced tapering (and the lateral or side etching induced) therein.

Feature profiles of poly-Si etched in Cl-2/O-2 plasmas have been analyzed through a mechanistic comparison between experiments and simulations. The emphasis was placed on a comprehensive understanding of the formation mechanisms for profile anomalies of tapering, microtrenching, and footing (or corner rounding near the feature bottom). Experiments were conducted in a commercial etching reactor with ultra-high-frequency plasmas by varying O-2 percentage, wafer stage temperature, rf bias power, and feed gas pressure. Simulations of the feature profile evolution were done by using a semiempirical, atomic-scale cellular model based on the Monte Carlo method that we have developed. The experiments indicated that sidewall profiles become more tapered with increasing O-2 addition to Cl-2 plasmas, while microtrenching and footing are pronounced in pure Cl-2 plasma, being suppressed with increasing O-2. A comparison with the simulations indicated that the tapered profiles are caused by the deposition of etch products/by-products on feature sidewalls from the plasma, being enhanced with increasing oxygen flux (due to synergistic effects between deposition of products/by-products and surface oxidation) and being reduced with increasing ion energy and neutral reactant flux. On the other hand, the footing is attributed to the redeposition of etch products on sidewalls from the feature bottom being etched, being reduced with increasing oxygen flux, ion energy, and neutral reactant flux. Microtrenching is caused by the ion reflection from feature sidewalls on incidence, being reduced with increasing oxygen flux (partly due to surface oxidation of the feature bottom) and being enhanced and then reduced with increasing ion energy and neutral reactant flux. The tapering, footing, and microtrenching were found to be closely related to each other: the footing near the feature bottom fades away under conditions of increased tapering of sidewalls, and the microtrenching is affected significantly by the degree of footing as well as the taper angle of the sidewalls.

A microplasma thruster of the electrothermal type has been developed with X-band (f = 11 GHz) microwaves and Ar as a propellant. The emphasis was placed on an understanding of distinguished features of the microplasma generation and thrust performance by X-band microwaves, compared with those by S-band (f = 4 GHz) ones. The thruster consisted of a microplasma source 2 mm in the inner diameter and 3–12 mm long with a rod antenna on the axis, followed by a converging-diverging micronozzle. Azimuthally symmetric surface wave-excited plasmas were established by microwaves at powers of ≤6 W, with the source pressure in the range 2–50 kPa at flow rates of 10–70 sccm. The plasma generation, nozzle flow, and thrust performance were numerically analyzed using a two-dimensional fluid model, coupled with an electromagnetic model for microwaves interacting with plasmas in the source region. Simulations indicated that higher frequency f = 11 GHz with the source chamber length Ls ≈ 1/4 of the driving wavelength λ is preferred for the microplasma thruster in terms of efficient plasma generation, gas heating, and thus thrust performance as well as system compactness; moreover, in f = 11 GHz discharges with longer Ls ≈ 3λ/4, standing-wave striation-like plasma structures occur in the axial direction. Experiments were made for f = 11 and 4 GHz microwaves with the respective Ls ≈ λ/4, where the plasma electron density and gas temperature in the microplasma source were measured by optical emission spectroscopy with a small amount of H2 and N2 added. The electron density and gas temperature were in the range of (4–12) × 1019 m−3 and 800–1000 K for f = 11 GHz, being ∼10%–50% higher than those for f = 4 GHz. The thrust performance was also measured by a target-type microthrust stand, giving a thrust, specific impulse, and thrust efficiency in the range 0.2–1.8 mN, 65–90 s, and 2%–14% for f = 11 GHz, which were ∼10%–15% higher than those for f = 4 GHz. These experimental results were consistent with those of simulations, depending on microwave frequency, power, chamber size, and gas flow rate.

Machine learning (ML) is now widespread. Traditional software engineering can be applied to the development ML applications. However, we have to consider specific problems with ML applications in therms of their quality. In this paper, we present a survey of software quality for ML applications to consider the quality of ML applications as an emerging discussion. From this survey, we raised problems with ML applications and discovered software engineering approaches and software testing research areas to solve these problems. We classified survey targets into Academic Conferences, Magazines, and Communities. We targeted 16 academic conferences on artificial intelligence and software engineering, including 78 papers. We targeted 5 Magazines, including 22 papers. The results indicated key areas, such as deep learning, fault localization, and prediction, to be researched with software engineering and testing.

Nanoscale surface roughening and ripple formation in response to ion incidence angle has been investigated during inductively coupled plasma etching of Si in Cl2, using sheath control plates to achieve the off-normal ion incidence on blank substrate surfaces. The sheath control plate consisted of an array of inclined trenches, being set into place on the rf-biased electrode, where their widths and depths were chosen in such a way that the sheath edge was pushed out of the trenches. The distortion of potential distributions and the consequent deflection of ion trajectories above and in the trenches were then analyzed based on electrostatic particle-in-cell simulations of the plasma sheath, to evaluate the angular distributions of ion fluxes incident on substrates pasted on sidewalls and/or at the bottom of the trenches. Experiments showed well-defined periodic sawtooth-like ripples with their wave vector oriented parallel to the direction of ion incidence at intermediate off-normal angles, while relatively weak corrugations or ripplelike structures with the wave vector perpendicular to it at high off-normal angles. Possible mechanisms for the formation of surface ripples during plasma etching are discussed with the help of Monte Carlo simulations of plasma-surface interactions and feature profile evolution. The results indicate the possibility of providing an alternative to ion beam sputtering for self-organized formation of ordered surface nanostructures.

Plasma-induced surface roughening and ripple formation has been studied based on Monte Carlo simulations of plasma-surface interactions and feature profile evolution during Si etching in Cl-based plasmas, with emphasis being placed on the role and effects of ion reflection from microstructural feature surfaces on incidence. The simulation model included the effects of Cl+ ion reflection (and/or its penetration into substrates) through calculating the momentum and energy conservation in successive two-body elastic collisions with substrate Si atoms every ion incidence. The “reflection coefficient ri” was then further introduced in the model (0 ≤ ri ≤ 1), representing the fraction of ions incident on surfaces with the reflection/penetration calculation scheme turned on. The coefficient ri is, in a sense, a measure of the reflection probability for impacts of an ion species onto Si surfaces relative to that for Cl+ impacts. Simulations for ion incidence angles of θi = 0°, 45°, and 75° onto substrate surfaces with incident energies in the range Ei = 20−500 eV showed that as ri is slightly decreased from unity, the roughness decreases substantially, and the ripple formation fades away: the roughness remains at the low level of stochastic roughening during etching for decreased ri ≤ ri* ≈ 0.95−0.75 (the critical ri* tends to be lower at higher Ei and θi) with no ripple structures at off-normal θi. This elucidates that the ion reflection is indispensable in surface roughening and rippling during plasma etching, and their degree relies significantly on the reflectivity of ions. Simulations further showed that at intermediate off-normal θi = 45°, the ripple wavelength increases significantly with decreasing ri, while the increase in amplitude is relatively less significant; thus, sawtooth-like ripple profiles pronounced for ri = 1 tend to be collapsed with decreasing ri. These effects of reduced ion reflection on plasma-induced surface roughening and ripple formation are discussed in terms of effectively enhanced smoothing due to neutral reactants, which competes with the roughening and rippling caused by ion bombardment.

We propose a new optical model for assigning the physical structure of plasma-damaged SiOC films examined by spectroscopic ellipsometry. A two-parameter Bruggeman’s effective medium approximation is used for estimating the thickness and volume fraction of a low-dielectric (ε) region (ε ∼ 1) in the SiO2 background. We introduced an optical model consisting of damaged and undamaged layers. The thickness and fraction of the damaged layer are fitted. Prediction was performed using this model for SiOC samples exposed to various plasmas, and the results were compared with those of scanning electron microscopy. We further applied this model to estimating the depth of damaged region in combination with a layer-by-layer wet-etching technique. In the case of He plasma exposure, the structural change induced by the damage extends 90–130 nm in depth. Since the degradation of interlayer dielectrics affects the circuit performance, the proposed optical model should be used for designing plasma processes.

Aggressive shrinkage and geometrical transition to three-dimensional structures in metal–oxide–semiconductor field-effect transistors (MOSFETs) lead to potentially serious problems regarding plasma processing such as plasma-induced physical damage (PPD). For the precise control of material processing and future device designs, it is extremely important to clarify the depth and energy profiles of PPD. Conventional methods to estimate the PPD profile (e.g., wet etching) are time-consuming. In this study, we propose an advanced method using a simple capacitance–voltage (C–V) measurement. The method first assumes the depth and energy profiles of defects in Si substrates, and then optimizes the C–V curves. We applied this methodology to evaluate the defect generation in (100), (111), and (110) Si substrates. No orientation dependence was found regarding the surface-oxide layers, whereas a large number of defects was assigned in the case of (110). The damaged layer thickness and areal density were estimated. This method provides the highly sensitive PPD prediction indispensable for designing future low-damage plasma processes.

Atomic- or nanometer-scale roughness on feature surfaces has become an important issue to be resolved in the fabrication of nanoscale devices in industry. Moreover, in some cases, smoothing of initially rough surfaces is required for planarization of film surfaces, and controlled surface roughening is required for maskless fabrication of organized nanostructures on surfaces. An understanding, under what conditions plasma etching results in surface roughening and/or smoothing and what are the mechanisms concerned, is of great technological as well as fundamental interest. In this article, we review recent developments in the experimental and numerical study of the formation and evolution of surface roughness (or surface morphology evolution such as roughening, smoothing, and ripple formation) during plasma etching of Si, with emphasis being placed on a deeper understanding of the mechanisms or plasma–surface interactions that are responsible for. Starting with an overview of the experimental and theoretical/numerical aspects concerned, selected relevant mechanisms are illustrated and discussed primarily on the basis of systematic/mechanistic studies of Si etching in Cl-based plasmas, including noise (or stochastic roughening), geometrical shadowing, surface reemission of etchants, micromasking by etch inhibitors, and ion scattering/chanelling. A comparison of experiments (etching and plasma diagnostics) and numerical simulations (Monte Carlo and classical molecular dynamics) indicates a crucial role of the ion scattering or reflection from microscopically roughened feature surfaces on incidence in the evolution of surface roughness (and ripples) during plasma etching; in effect, the smoothing/non-roughening condition is characterized by reduced effects of the ion reflection, and the roughening-smoothing transition results from reduced ion reflections caused by a change in the predominant ion flux due to that in plasma conditions. Smoothing of initially rough surfaces as well as non-roughening of initially planar surfaces during etching (normal ion incidence) and formation of surface ripples by plasma etching (off-normal ion incidence) are also presented and discussed in this context.

We proposed an electrical evaluation method on the basis of leakage current–stress time (I–t) measurement for plasma-induced damage (PID) in interlayer dielectric films. The I–t measurement, which is widely used for the time-dependent dielectric breakdown (TDDB) test is applied to the evaluation of PID in thick SiO2 films. We focus on the electron trapping process during the TDDB test. The initial stage of time evolution of leakage current is monitored, and compared among the samples damaged under various plasma conditions. It was found that the initial electron trapping rate depends on the amount of damage, i.e., the rate increases with increasing in process time (tpr) and dc self-bias voltage (Vdc). The proposed technique can be used to assess the reliability degradation of interlayer dielectric films caused by PID.

Effects of initial roughness on the evolution of plasma-induced surface roughness have been investigated during Si etching in inductively coupled Cl2 plasmas, as a function of rf bias power or ion incident energy in the range Ei ≈ 20–500 eV. Experiments showed that smoothing of initially rough surfaces as well as non-roughening of initially planar surfaces can be achieved by plasma etching in the smoothing mode (at high Ei) with some threshold for the initial roughness, above which laterally extended crater-like features were observed to evolve during smoothing. Monte Carlo simulations of the surface feature evolution indicated that the smoothing/non-roughening is attributed primarily to reduced effects of the ion scattering or reflection from microscopically roughened feature surfaces on incidence.

We propose an electrical method, named capacitance–voltage ( C – V ) monitoring, for quantifying plasma-induced damage (PID) to interlayer dielectrics. By this method, we measure the C – V hysteresis loops to assign carrier trap sites created by PID, and simultaneously obtain the change in the dielectric constant and thickness. We optimized the bias-sweep configuration for measuring the hysteresis curves. It is found that the C – V curve shifted in the negative direction during the optimized voltage sweep from accumulation to inversion in a pseudo-metal–oxide–semiconductor (MOS) structure. This implies the appearance of net positively charged sites owing to PID, presumably near the surface of the SiOC film. We estimate the density of defects created near the surface by monitoring the obtained C – V hysteresis curve shift. Since the degradation of interlayer dielectrics affects the circuit performance, the proposed quantitative method should be used for plasma process designs.