We report the effect of N2 addition to C4F8 and C4F8/Ar discharges on plasma etching rates of organosilicate glass (OSG) and etch stop layer materials (Si3N4 and SiC), and the results of surface chemistry studies performed in parallel. N2 addition exhibits different effects in C4F8 and C4F8/Ar plasmas, which may be explained by a higher plasma density, electron temperature, and possibly, the presence of argon metastable species in the C4F8/Ar plasma, all of which serve to dissociate N2 more effectively. When N2 is added to a C4F8/Ar plasma, a reduction of the steady-state fluorocarbon surface layer thickness, one of the key parameters that controls the etching rate and etching selectivity on partially etched samples, is observed. This effect leads to a loss of etching selectivity for C4F8/Ar/N2 discharges. Adding N2 to C4F8 plasmas without Ar enhances the steady-state fluorocarbon layer thickness. X-ray photoelectron spectroscopy analysis shows, in this case, that there is an important change in the stoichiometry of either passively deposited films or the fluorination reaction layers formed on etching samples: A significant amount of nitrogen is incorporated in the fluorocarbon film for deposited films, which implies that CxNy needs to be removed to achieve an etching condition. The incorporation of nitrogen in fluorocarbon films could reduce the etchant supply for Si3N4, or OSG, from the gas phase, especially for C4F8/Ar/N2 plasmas, but not for SiC owing to the differences of the chemical compositions. SiO2 and Si are also studied for comparison materials. The etching behavior of SiO2 is similar to that of OSG and Si3N4, while Si behaves more similar to SiC during fluorocarbon etching. In addition, a comparison of N2 and O2 addition to C4F8 or C4F8/Ar plasma in terms of consequences on etching behavior of the aforementioned materials is presented.

The role of ion bombardment in plasma etch selectivity of organosilicate glass to etch stop layers of silicon carbide and silicon nitride has been investigated in a C4F8/N2/Ar plasma using a method that produces a narrow ion energy distribution (IED) at the substrate surface. The effects of the narrow IED are compared with those of the broad, bimodal IED produced by the conventional sinusoidal bias voltage wave form (at 13.56 MHz). A comparison of etch rate versus average ion bombardment energy shows a higher ion energy threshold for etching, a larger gap between the thresholds for the two materials, and high selectivity over a wider range of bias voltage with the narrow IED. A physical explanation of the observed phenomena is proposed.

An in-line mass spectrometer and Langmuir probes have been employed to examine mechanisms of plasma immersion hydrogen passivation of grain boundary defects in polycrystalline silicon thin film transistors. Relative fluxes of H ϩ and H 2 ϩ as well as total ion current density were measured at the substrate location in an electron cyclotron resonance hydrogen discharge. Measurements were made over a range of operating conditions over which passivation rates have been shown to vary dramatically. Data presented show a strong correlation of both H ϩ flux and ion bombardment energy with good transistor performance obtained at operating pressures below 1 mTorr. This suggests that discharge operating conditions that promote dissociation of H 2 to form H and H ϩ ͑which may diffuse more rapidly through solid material than H 2 ͒, as well as increased sheath voltages and therefore ion energy at the substrate, are important to obtaining acceptable process rates. © 1995 American Institute of Physics.Acceptable electrical performance of polysilicon thin film transistors ͑TFTs͒ for several applications, including driver circuitry for active matrix liquid crystal displays ͑AMLCD͒, requires the hydrogen passivation of defect sites in the polycrystalline silicon grain boundaries.1,2 One of the most promising methods for passivation is exposure of the fabricated transistors to low pressure ͑Ͻ1 mTorr͒ hydrogen plasmas generated in an electron cyclotron resonance ͑ECR͒ source. However, ECR-based reactors may not be well suited for large area substrate processing under development for AMLCD flat panel displays. To aid in the design of processes based on more suitable reactor types employing, for example, inductively coupled or helicon plasmas, it is helpful to understand the basic mechanisms responsible for efficient plasma exposure hydrogenation. Specifically, we address the hypothesis that high passivation rates observed in low pressure ECR hydrogen discharges are the result of enhanced dissociation of H 2 at discharge pressures below 1 mTorr.3 It is believed that because atomic hydrogen diffuses through solid material more readily than molecular hydrogen, conditions that favor a high flux of either neutral or ionized atomic hydrogen to the substrate will result in high rates of hydrogen passivation.To investigate these mechanisms, we examine plasma conditions in a H 2 ECR discharge identical to that used in a study of polysilicon passivation, over a range of operating parameters over which passivation effectiveness was found to vary dramatically.1 This study involves examination of the mass distribution and flux of ions, but not neutrals, incident on the substrate. There are practical reasons for studying only the ions, and it is believed that the ion data on its own provides substantial support for the significance of the role of atomic hydrogen. First, the presence of atomic hydrogen ions, H ϩ , can be considered to be a good indicator of the presence of neutral atomic hydrogen. In addition, even if the concentration of H ϩ ions...

Articles you may be interested inMechanism of selective SiO2/Si etching with fluorocarbon gases (CF4, C4F8) and hydrogen mixture in electron cyclotron resonance plasma etching system

Mask undercut in the time-multiplexed deep silicon etch process is becoming an increasingly significant issue as it is used to produce smaller critical dimension features. Models of the process must contain the necessary physics to reproduce the dependencies of mask undercut. We argue that the reason undercut develops is the dependence of the deposition step on ion flux. Our experiments of C4F8 (and CHF3 not shown) plasmas show that the film growth is dominantly ion-enhanced. This leads naturally to a mask undercut that increases in time. A more neutral flux dominant deposition step would result in reduced mask undercut.